Nociceptive sensory neurons convey pain signals to the CNS using action potentials. Loss-of-function mutations in the voltage-gated sodium channel NaV1.7 cause insensitivity to pain (presumably by reducing nociceptor excitability) but efforts to treat pain by inhibiting NaV1.7 pharmacologically have largely failed. This may reflect the variable contribution of NaV1.7 to nociceptor excitability. Contrary to claims that NaV1.7 is necessary for nociceptors to initiate action potentials, we show that nociceptors can achieve equivalent excitability using different combinations of NaV1.3, NaV1.7, and NaV1.8. Selectively blocking one of those NaV subtypes reduces nociceptor excitability only if the other two subtypes are weakly expressed. For example, excitability relies on NaV1.8 in acutely dissociated nociceptors but responsibility shifts to NaV1.7 and NaV1.3 by the fourth day in culture. A similar shift in NaV dependence occurs in vivo after inflammation, impacting ability of the NaV1.7-selective inhibitor PF-05089771 to reduce pain in behavioral tests. Flexible use of different NaV subtypes – an example of degeneracy – compromises the reliable modulation of nociceptor excitability by subtype-selective inhibitors. Identifying the dominant NaV subtype to predict drug efficacy is not trivial. Degeneracy at the cellular level must be considered when choosing drug targets at the molecular level.
+
+SIGNIFICANCE STATEMENT
+
Nociceptors can achieve equivalent excitability using different sodium channel subtypes. The analgesic efficacy of subtype-selective drugs hinges on which subtype controls excitability. This contingency likely contributes to poor clinical outcomes.
+
+
+
+
+
+
+
+
+Competing Interest Statement
SAP is on the Scientific Advisory Boards of Boston Scientific and Presidio Medical and has received grant funding from Boston Scientific.
+
+Summary of Updates:
+
This version has been revised through addition of new text to the Introduction and Discussion. Formatting has also been adjusted, which includes changes to the organization of figures. The results themselves have not changed.
+
+
+
+
+
+INTRODUCTION
+
Chronic pain affects between 11 and 40% of the population worldwide (1). Neuropathic pain, which is pain arising from damage to the somatosensory nervous system, is particularly hard to treat with only 30% of patients achieving moderate (≥30%) relief using available treatments (2, 3). New treatments are needed but a meagre 11% of analgesic drugs entering phase 1 trials are ultimately approved (4), which has triggered debate about why basic science discoveries are not yielding improved clinical outcomes (5). Suggested explanations include flaws in preclinical animal testing (6, 7) or clinical trial design (8) but biological explanations must also be considered. For example, degeneracy – the ability of a biological system to achieve equivalent function using different components (9) – complicates modulation of neuronal excitability by allowing changes in diverse ion channels to potentially subvert the therapeutic effect of a drug targeting a particular channel (10).
+
Like most neurons, nociceptive sensory neurons (nociceptors) rely on spikes to transmit information. Their excitability is thus critical for relaying information to the CNS. Nociceptor excitability is increased in many pathological pain conditions and the resultant increase in afferent input drives chronic pain (11–13). Neuronal excitability depends on the complex interplay between diverse ion channels (14–16) but some channels seem to be particularly important for pain. For instance, loss- or gain-of-function mutations in the gene SCN9A, which encodes the voltage-gated sodium channel NaV1.7, cause congenital insensitivity to pain (CIP) or painful neuropathies, respectively (17-19; for review see 20). In rodents, nociceptor-specific deletion of NaV1.7 abolishes acute and inflammatory pain (21) but not neuropathic pain (22, 23). Neuropathic pain is blocked by deleting NaV1.7 globally, including from sympathetic neurons (24, 25), though not if the deletion is induced in adulthood (26). Furthermore, loss-of-function mutations in NaV1.7 do not consistently reduce nociceptor excitability (see Discussion) and the associated insensitivity to pain involves increased opioid signaling (27, 28), consistent with naloxone’s ability to restore pain sensitivity in CIP patients (27, 29). These observations cast doubt on whether NaV1.7 mutations produce CIP by reducing nociceptor excitability, pointing instead to a less direct mechanism that may be harder to reproduce pharmacologically.
+
Notwithstanding such reservations, several NaV1.7-selective drugs have been developed (30–32) but none have yet passed phase 2 clinical trials (33–36). This has been attributed to poor target engagement (35, 37–39) yet prevention of the flare response by PF-05198007, a NaV1.7-selective inhibitor, argues that at least some NaV1.7 channels are blocked (40). But CIP patients exhibit a normal flare response (41), suggesting that their C fibers compensate for chronic loss of NaV1.7 channels. Other NaV1.7-selective inhibitors have struggled in phase 1 trials because of autonomic side effects (e.g. 42), as might be expected if those drugs block NaV1.7 channels on sympathetic neurons, which is apparently necessary to prevent/reverse neuropathic pain (see above). But CIP patients exhibit normal autonomic function (17, 41), suggesting that their sympathetic neurons also compensate for chronic loss of NaV1.7 channels. In those patients, might similar compensation occur in nociceptors and restore pain, only for that effect to be masked by enhanced opioid signaling (see above)? Descriptions of NaV1.7 as “the” threshold channel imply that it is irreplaceable for nociceptor excitability, consistent on the surface with pain insensitivity due to loss-of-function mutations in NaV1.7 but inconsistent with some past electrophysiological data (43, 44). Clarifying whether nociceptors rely on NaV1.7 is an unresolved issue important for predicting the analgesic efficacy of NaV1.7-selective inhibitors.
+
A serendipitous observation prompted us to reassess the role of NaV1.7 in nociceptor excitability and the implications for drug efficacy. Specifically, we observed that tetrodotoxin (TTX), which inhibits NaV1.7 and several other TTX-sensitive (TTX-S) sodium channels, had variable effects in nociceptors, dramatically reducing their excitability in some conditions but not in others. This variability reveals that nociceptors can achieve equivalent excitability using different sodium channel subtypes, some of which are TTX-resistant (TTX-R). We demonstrate that a NaV1.7-selective inhibitor produces analgesia only when nociceptor excitability relies on NaV1.7. Insofar as increasingly selective drugs are more likely to have their efficacy subverted by degeneracy, our results have profound yet underappreciated implications for target selection and drug development.
+
+
+RESULTS
+
+Equivalent excitability can arise from different voltage-gated sodium (Na<sub>V</sub>) channel subtypes
+
Small dorsal root ganglion (DRG) neurons (soma diameter <25 µm) tend to spike repetitively when depolarized by current injection (45). In our sample, most small neurons genetically identified as nociceptors (see Methods) spiked repetitively when tested 2-8 hours after dissociation (DIV0) or after 4-7 days in culture (DIV4-7), though the proportion of repetitively spiking neurons increased slightly over that interval (ξ2=4.51, p=0.034, chi-square test) (Fig. 1A). Strikingly, 100 nM TTX had no effect on the spiking pattern at DIV0 but converted all but one neuron to transient spiking at DIV4-7. Amongst neurons that spiked repetitively at baseline, TTX reduced the firing rate and increased rheobase only at DIV4-7 (Fig. 1B). TTX reduced spike height at DIV0 and DIV4-7, but more so at DIV4-7. There was a significant increase in capacitance and leak conductance density between DIV0 and DIV4-7, but no change in resting membrane potential (Fig. 1C). Normalizing leak conductance by capacitance (which increases over time because of neurite growth) disambiguates whether changes in input resistance reflect changes in cell size or membrane leakiness. Consistent with current clamp data, voltage clamp recordings showed that only a small fraction of sodium current is TTX-S at DIV0, whereas nearly all sodium current was blocked by TTX at DIV4-7 (Fig. 1D). Previous studies suggested that TTX-R channels play an important role in nociceptor excitability (46–48). Our initial results confirm this for DIV0 but show that their contribution diminishes after a few days in culture, with TTX-S channels becoming dominant by DIV4. Despite this reconfiguring of NaV channels, excitability was remarkably stable, consistent with previous work showing little change in excitability after axotomy despite large (but evidently counterbalanced) changes in TTX-R and TTX-S currents (43, 49). We show later that similar changes develop in vivo following inflammation with consequences for drug efficacy assessed behaviorally (see Fig. 8), meaning the NaV channel reconfiguration described above is not a trivial epiphenomenon of culturing.
+
+
+
Different Na<sub>V</sub> subtypes produce equivalent excitability at different days in vitro (DIV).
(A) Representative responses of small DRG neurons to current injection at rheobase and 3x rheobase when tested on DIV0 (blue) or DIV4-7 (red) before (dark) and after (pale) bath application of 100 nM TTX. At DIV0, TTX did not alter spiking pattern (ξ2=0.25, p=0.617, McNemar test) or significantly reduce firing rate (F1,72=1.527, p=0.24, two-way repeated measure (RM) ANOVA; n=13). At DIV4-7, TTX significantly altered spiking pattern, converting all but one neuron to transient spiking (ξ2=20.05, p<0.0001), and it significantly reduced firing rate (F1,132=43.157, p<0.001, n=23). Only neurons with repetitive spiking at baseline are included in the firing rate plot. (B) At DIV0, TTX did not affect rheobase (Z24=1.129, p=0.265, Wilcoxon rank test) but did reduce spike height (T24=3.092, p=0.005, paired t-test). At DIV4-7, TTX increased rheobase (Z28=4.681, p<0.001, Wilcoxon rank test) and dramatically reduced spike height (T28=20.333, p<0.001, paired t-test). Notably, neurons at DIV0 and DIV4-7 did not differ in their baseline rheobase (U=316, p=0.425, Mann-Whitney test) or spike height (T52=0.322, p=0.749, t-test). (C) Neurons at DIV0 and DIV4-7 differed in their total capacitance (T52=6.728, p<0.001, t-test) and leak conductance density (U=216, p=0.011, Mann-Whitney test) but not in their resting membrane potential (T52=1.668, p=0.101, t-test). (D) Sample voltage clamp recordings with command voltage stepped from -85 mV to +15 mV in 5 mV increments, before and after TTX. Sodium current was not significantly reduced by TTX at DIV0 (F1,72=3.585, p=0.107, two-way RM ANOVA; n=7 neurons) but was completely abolished by TTX at DIV4-7 (F1,108=33.526, p<0.001; n=10 neurons). *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc tests in A and D.
+
+
+
+
+Different Na<sub>V</sub> channel subtypes control nociceptor excitability at DIV0 and DIV4-7
+
Next, we sought to identify the NaV subtype responsible for repetitive spiking at each time point, starting with DIV0. Of the TTX-R NaV channels expressed by nociceptors, NaV1.8 has been implicated in repetitive spiking (47, 48). We measured sodium current in voltage clamp before and after applying the NaV1.8-selective inhibitor PF-01247324 (PF-24) (50). At DIV0, 1 µM PF-24 abolished most of the sodium current (Fig. 2A). The PF-24-sensitive current had slow inactivation kinetics, like the TTX-R current and unlike the fast TTX-S current in Figure 1D, and consistent with previous descriptions of NaV1.8 (51). A different Nav1.8 antagonist, A-803467, had similar effects (Fig. 2 – figure supplement 1). In current clamp, PF-24 converted 7 of 8 repetitively spiking neurons to transient spiking and significantly reduced evoked spiking (Fig. 2B). It also increased rheobase and decreased spike height but did not affect resting membrane potential (Fig. 2C). PF-24 had negligible effects when tested at DIV4-7 (Fig. 2 – figure supplement 2). These results show that NaV1.8 is the predominant NaV subtype at DIV0 and is necessary for repetitive spiking at that time point. To test the sufficiency of NaV1.8 to produce repetitive spiking, we tuned a conductance-based model neuron (see Methods) to reproduce DIV0 data described above. In this DIV0 model, inclusion of NaV1.8 conductance was sufficient to generate repetitive spiking (Fig. 2D left). The necessity of NaV1.8 for repetitive spiking at DIV0 was also recapitulated: 85% reduction in the NaV1.8 conductance converted spiking from repetitive to transient (Fig. 2D and Supplementary Table 1).
+
+
+
Na<sub>V</sub>1.8 is necessary for repetitive spiking at DIV0.
(A) Sample voltage clamp recordings show that sodium current was almost completely abolished by the NaV1.8 inhibitor PF-24 (1 µM). Peak current was significantly reduced by PF-24 (F1,72=12.651, p<0.012, two-way RM ANOVA; n=7). Another NaV1.8 inhibitor, A-803467, had a similar effect (see Fig. 2 – figure supplement 1). (B) PF-24 significantly altered spiking pattern (χ2=5.14, p=0.0233, McNemar test) and reduced firing rate (F1,42=11.946, p=0.011, two-way RM ANOVA; n=8). (C) PF-24 significantly increased rheobase (Z15=2.783, p=0.003, Wilcoxon rank test) and reduced spike height (T15=3.151, p=0.007, paired t-test) but did not affect resting membrane potential (T15=0.304, p=0.765, paired t-test). PF-24 had limited effects at DIV4-7 (Fig 2 – figure supplement 2). (D) A computational model reproduced the effect of NaV1.8 on spiking pattern (also see Supplementary Table 1). The PF-24 effect was simulated as a ∼85% reduction in NaV1.8 (𝑔̄Nav1.8== 4mS/cm2). *, p<0.05; **; p<0.01; Student-Newman-Keuls post-hoc tests in A and B.
+
+
+
Next, we sought to identify the NaV subtype responsible for repetitive spiking at DIV4-7 using PF-05089771 (PF-71) to inhibit NaV1.7 (40, 52) and ICA-121431 (ICA) to inhibit NaV1.1/1.3 (53, 54). In voltage clamp, sodium current was significantly reduced by 30 nM PF-71, and most of the remaining current was blocked by 1 µM ICA (Fig. 3A). In current clamp, each inhibitor (applied separately) converted a significant proportion of neurons to transient spiking and significantly reduced firing rate (Fig. 3B). This argues that NaV1.7 and NaV1.1/1.3 are both necessary for repetitive spiking at DIV4-7. Inhibiting NaV1.7 increased rheobase, unlike inhibiting NaV1.1/1.3, and caused a stronger reduction in spike height (Fig. 3C). Neither affected resting membrane potential. These results show that NaV1.7 is the predominant NaV subtype at DIV4-7, but not the only one. PF-71 had negligible effects when tested at DIV0 (Fig. 3 – figure supplement 1). We re-tuned our computational model to reproduce DIV4-7 data, with both NaV1.7 and NaV1.3 being required to produce repetitive spiking, meaning neither channel is individually sufficient (Fig. 3D and Supplementary Table 1). That said, inserting a higher density of either NaV1.7 or NaV1.3 could produce repetitive spiking in the absence of the other subtype (Fig. 3 – figure supplement 2), consistent with NaV1.7 and NaV1.3 also being interchangeable.
+
+
+
Na<sub>V</sub>1.3 and Na<sub>V</sub>1.7 are necessary for repetitive spiking at DIV4-7.
(A) Sample voltage clamp recordings show that sodium current was reduced by the NaV1.7 inhibitor PF-71 (30 nM) and by the NaV1.1/1.3 inhibitor ICA (1 µM). Peak current was significantly reduced by PF-71 and ICA (F2,192=26.361, p<0.001, two-way RM ANOVA; n=9). (B) PF-71 and ICA both significantly altered spiking pattern (χ2=4.17, p=0.041 and χ2 =7.11, p=0.0077, respectively, McNemar tests) and significantly reduced firing rate (F1,54=40.659, p<0.001, n=10 and F1,78=35.156, p<0.001, n=14, respectively, two-way RM ANOVAs). (C) PF-71 significantly increased rheobase (Z18=3.464, p<0.001, Wilcoxon rank test) and decreased spike height (T18=7.946, p<0.001, paired t-test). ICA did not significantly alter rheobase (Z18=1.248, p=0.225) but did reduce spike height (T18=3.243, p=0.005). Neither drug affected resting membrane potential (T15=1.681, p=0.113 for PF-71; T18=-1.132, p=0.272 for ICA, paired t-test). PF-71 had negligible effects at DIV0 (Fig. 3 – figure supplement 1). (D) A computational model reproduced the combined effects of NaV1.3 and NaV1.7 on spiking pattern (also see Supplementary Table 1 and Fig. 3 – figure supplement 2). PF-71 effect was simulated as a 70% reduction in NaV1.7 (𝑔̄Nav1.7= 10.5 mS/cm2). ICA effect was simulated as a 90% reduction in NaV1.3 ((𝑔̄Nav1.3= 0.035 mS/cm2). *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc tests in A and B.
+
+
+
+
+Acutely interchanging Na<sub>V</sub> subtypes does not affect spiking pattern
+
The ability of NaV1.3, NaV1.7 and NaV1.8 to each encourage repetitive spiking is seemingly inconsistent with the common view that each NaV subtype contributes selectively to a different phase of the spike (for example, Figure 3 in ref 55). If NaV1.8 were to activate exclusively at suprathreshold voltages, it could not initiate spikes and a different perithreshold-activating NaV channel would be needed, which is clearly inconsistent with our data. To verify that NaV1.7 and NaV1.8 currents are each sufficient to produce repetitive spiking, we tested whether the NaV1.8 current necessary for spiking in our DIV0 model could be replaced with NaV1.7, and whether the NaV1.7 current necessary for spiking in our DIV4-7 model could be replaced with NaV1.8. In both cases, repetitive spiking was restored after inserting the alternate current (Fig. 4A). We then proceeded with equivalent experiments in real neurons, inhibiting NaV1.8 with PF-24 on DIV0 or NaV1.7 with PF-71 on DIV4-7, and then introducing the alternate channel virtually using dynamic clamp (see Methods). The replacement was successful in all neurons tested (Fig. 4B). Inserting virtual NaV1.8 after inhibiting native NaV1.8 also restored repetitive spiking, and likewise for NaV1.7 (Fig. 4 – figure supplement 1), verifying that our virtual channels were equivalent to the native channels we aimed to replace. Apart from maximal conductance density, which was titrated in each neuron, all other parameters used for dynamic clamp were identical to simulations. The success of dynamic clamp experiments helps validate our computational models insofar as virtual NaV1.7 and NaV1.8 currents interacted appropriately with native currents to produce repetitive spiking in real neurons, the same way they interact with other simulated currents in the model neuron.
+
+
+
Na<sub>V</sub>1.7 and Na<sub>V</sub>1.8 are each sufficient to produce repetitive spiking in DIV0 and DIV4-7 neurons.
(A) The computational model predicts that the NaV1.8 conductance, which is “necessary” for repetitive spiking at DIV0 can, in principle, be replaced by NaV1.7 (left), and vice versa at DIV4-7 (right). (B) Replacement experiments involved inhibiting native channels pharmacologically and then introducing virtual conductances using dynamic clamp. At DIV0 (left), inhibiting native NaV1.8 (with PF-24) converted neurons to transient spiking, but introducing virtual NaV1.7 reverted neurons to repetitive spiking (in 3 of 3 neurons tested). At DIV4-7, inhibiting native NaV1.7 (with PF-71) converted the neuron to transient spiking, but introducing virtual NaV1.8 reverted neurons to repetitive spiking (in 4 of 4 neurons tested). Repetitive spiking was likewise restored by replacing the blocked native channel with the corresponding virtual channel (Fig. 4 – figure supplement 1). Parameters for virtual channels were identical to simulations except for the maximal conductance density, which was titrated in each cell.
+
+
+
With the model neurons thus validated, we used simulations to measure NaV1.7 and NaV1.8 currents during different phases of the spike (Fig. 5A-D). Since inward (depolarizing) current at voltages near spike threshold is critical for spike initiation (56; see above), we sought to identify which NaV contributes to that current. In the DIV0 model, NaV1.7 dominated during onset of the first spike but all subsequent spikes were initiated by NaV1.8 (Fig. 5A,B). This occurred because the small NaV1.7 conductance at DIV0 quickly inactivated during the first spike and remained inactive during subsequent spikes (Fig. 5 – figure supplement 1A). This is consistent with experimental results, where repetitive spiking at DIV0 was unaffected by inhibiting NaV1.7 (see Fig. 1 and Fig. 3 – figure supplement 1) but was prevented by inhibiting NaV1.8 (see Fig. 2). However, inactivation of NaV1.7 after the first spike was reflected by an increase in voltage threshold between the first and second spike, both in the model (Fig. 5A) and in experiments (Fig. 5E). This unexpected simulation result also predicted that TTX should affect the voltage threshold of the first spike in DIV0 neurons despite not having other notable effects (see Fig. 1); as predicted, TTX caused a significant depolarizing shift in voltage threshold at DIV0 (Fig. 5 – figure supplement 2), further validating our model. In the DIV4-7 model, NaV1.7 and NaV1.3 contributed to initiation of all spikes whereas NaV1.8 was negligible (Fig. 5C,D). Even though inactivation reduced NaV1.7 and NaV1.3 current after the first spike (Fig. 5 – figure supplement 1B), those channels nonetheless provided sufficient inward current to support repetitive spiking at DIV4-7. Inactivation at DIV4-7 was reflected, however, in a combination of higher threshold and lower spike overshoot for the second spike, both in the model (Fig. 5C,D) and in experiments (Fig. 5E). These results demonstrate that each NaV subtype does not contribute exclusively to a particular phase of the spike, and nor is each spike phase mediated exclusively by a particular NaV subtype; instead, each subtype contributes preferentially to a different spike phase depending on its voltage-dependency, but conductance density and inactivation status are both important. Indeed, an subtype’s contribution can shift rapidly (because of channel inactivation) or slowly (because of gene expression changes).
+
+
+
Contribution of Na<sub>V</sub>1.7 and Na<sub>V</sub>1.8 to spike initiation in DIV0 and DIV4-7 neurons.
(A) Voltage (top) for first (left) and second (right) spikes in the DIV0 model aligned with voltage activation curves for each NaV subtype (bottom). Dashed line shows voltage threshold (defined as V where dV/dt reaches 5 mV/ms). (B) Conductance plotted against voltage to create a phase portrait (top) showing NaV conductance at different phases of the spike. Inset shows full voltage range; main graph zooms in on voltages near threshold. Bottom plots show current plotted over the same voltage range. Whereas NaV1.7 (orange) mediated nearly all perithreshold inward current for the first spike, voltage threshold increased – because NaV1.7 inactivated (Fig. 5 – figure supplement 1) – and NaV1.8 (green) mediated nearly all perithreshold inward current for the second spike. The unexpected contribution of NaV1.7 to the first spike correctly predicted that TTX increases voltage threshold in DIV0 neurons (Fig. 5 – figure supplement 2). (C, D) In the DIV4-7 model, NaV1.7 (orange) and NaV1.3 (maroon) contributed to initiation of all spikes whereas the contribution of NaV1.8 was negligible (due entirely to its low expression level). (E) Sample experimental traces showing differences in the first (blue/red) and second (grey) spikes at DIV0 and DIV4-7. Plots summarize differences (ý) in threshold, overshoot potential, and spike rise time between 1st and 2nd spikes during repetitive spiking evoked by current injection. At DIV0, the 1st and 2nd spikes differ significantly in their threshold (T8=2.522, p=0.036, one-sample t-test) and overshoot (T8=0.038, p=0.038) but not rise time (T8=0.249, p=0.810). At DIV4-7, the 1st and 2nd spikes differ in all measures (threshold: T7=7.613, p<0.001; overshoot: T7=-9.849, p<0.001; rise time: T7=5.979, p<0.001). Statistical results (green) show that differences between 1st and 2nd spike at DIV4-7 are significantly larger than differences at DIV1 (threshold: T15=-3.847, p=0.002; overshoot: T15=7.922, p<0.001; rise time: T15=-5.617, p<0.001, unpaired t-tests), consistent with our computational model.
+
+
+
+
+Control of changes in Na<sub>V</sub> subtype expression between DIV0 and DIV4-7
+
Next, we sought to identify the basis for the slow shift in which NaV subtype controls nociceptor excitability. Figure 6A shows mRNA levels for NaV1.7 and NaV1.8 relative to a housekeeping gene (left) and to each other (right). NaV1.7 mRNA levels exceeded NaV1.8 mRNA levels at both at both DIV0 and DIV7. Both decreased between DIV0 and DIV7, but NaV1.8 more so, resulting in a significant decrease in the NaV1.8:NaV1.7 mRNA ratio. This pattern is consistent with the reduced role of NaV1.8 at DIV4-7 but is inconsistent with the negligible role of NaV1.7 at DIV0; specifically, we expected NaV1.7 mRNA levels to increase between DIV0 and DIV7. Next, we investigated if functional changes were better reflected by changes in protein levels. Immunofluorescence for NaV1.8 was higher than for NaV1.7 at DIV0, and that ratio reversed at DIV7 (Fig. 6B), consistent with functional changes. Moreover, cercosporamide (10 µM), a potent inhibitor of the eukaryotic translation Initiation Factor 4E (eIF4E), significantly mitigated the decrease in NaV1.8 immunofluorescence and the increase in NaV1.7 immunofluorescence when applied to cultured neurons for 24 or 120 hours prior to measurements on DIV5 (Fig. 6C). Beyond showing that their mRNA levels do not correlate well with NaV contributions to nociceptor excitability, reminiscent of some previous work (e.g. 57), these results suggest that translational regulation is crucial, though membrane trafficking and other downstream processes likely also contribute (58, 59).
+
+
+
Protein levels, but not mRNA, reflect functional contributions of Na<sub>V</sub> subtypes at DIV0 and DIV7.
(A) Both NaV1.8 and NaV1.7 mRNA levels (relative to a housekeeping gene (HKG), see Methods) decreased significantly between DIV0 and DIV4-7 (factor 1: time, F1,12=56.677, p<0.001, factor 2: subtype, F1,12=17.952, p=0.001, two-way ANOVA and Student-Newman-Keuls post-hoc tests on log transformed data, n=4 mice per time point) but more so for NaV1.8 than for NaV1.7 (interaction: time x subtype, F1,12= 11.455, p=0.005). The differential reduction yielded a significantly higher NaV1.8: NaV1.7 ratio at DIV0 than at DIV7 (T6=21.375, p<0.001, unpaired t-test) but the increasing functional contribution of NaV1.7 between DIV0 and DIV4-7 remains unaccounted for. (B) Immunoreactivity (IR) for NaV1.8 protein exceeded NaV1.7-IR at DIV0, but the opposite was true on DIV4-7, consistent with the functional contribution of each subtype. NaV-IR was measured relative to YFP intensity in the same cell, and then each cell’s NaV1.8:YFP ratio was considered relative to the average NaV1.7:YFP ratio in the co-processed coverslip (left) or average NaV1.8:YFP ratio was considered relative to the average NaV1.7:YFP ratio in the same animal (right). Ratios were >1 at DIV0 but decreased significantly at DIV4-7 (U=78, p<0.001, n=37 for DIV0, n=40 for DIV4-7, Mann-Whitney test (left) and T6=4.046, p=0.007, unpaired t-test (right)). (C) Chronically applied cercosporamide (10 µM) mitigated the changes in NaV1.8-and NaV1.7-IR at DIV5 (NaV1.8: H3=157.95, p<0.001; NaV1.7: H3=80.662, p<0.001; One-way ANOVA on ranks, Dunn’s post-hoc tests, p<0.05 for all pairs). Panel on the right shows data normalized to baseline (DIV0) to emphasize relative changes.
+
+
+
+
+Analgesic efficacy of subtype-selective drugs depends on which Na<sub>V</sub> controls nociceptor excitability
+
If a NaV1.7-selective inhibitor mediates analgesia by modulating nociceptor excitability, its analgesic efficacy hinges on nociceptor excitability being controlled by NaV1.7. Accordingly, we predicted that the NaV1.7-selective inhibitor PF-71 would have little if any effect on paw withdrawal under normal conditions, when NaV1.8 controls nociceptor excitability (Fig. 2 and Fig. 3 – figure supplement 1), but would be effective if NaV1.7 took over control. Inflammation increases NaV1.7 channel trafficking and membrane expression (60–63). To test if inflammation increased NaV1.7’s influence on nociceptor excitability, we recorded neurons acutely dissociated (DIV0) from DRGs of mice whose hind paw was injected with CFA three days prior. Inflammation caused nociceptors to become much more variable in their reliance of specific NaV subtypes (Fig. 7A). Despite this variability, inhibiting NaV1.7 with PF-71 converted a significantly higher proportion of neurons to transient spiking after CFA (42%) than in control neurons (0%) (Fig. 7B, left); subsequent application of PF-24 to inhibit NaV1.8 converted just 14% of CFA neurons to transient spiking vs 88% of control neurons (Fig 7B, right). PF-71 also significantly affected resting membrane potential, rheobase, and spike height after CFA (Fig. 7C), unlike in control neurons (see Fig. 3 – figure supplement 1).
+
+
+
Inflammation alters Na<sub>V</sub> subtype contribution to nociceptor excitability.
(A) Sample responses in DIV0 neurons from mice injected with CFA three days earlier. In 12 cells tested, PF-71 converted 5 neurons to transient spiking (i), encouraged repetitive spiking in 4 neurons (ii), and had no effect in 3 neurons (iii), thus highlighting increased heterogeneity after CFA. (B) At DIV0, the effect of PF-71 differed significantly between CFA and control neurons, converting 42% (5 of 12) CFA neurons from repetitive to transient spiking vs 0% (0 of 9) control neurons (p=0.0451, Fisher Exact test). Applying PF-24 to neurons that continued to spike repetitively after PF-71 had little effect on CFA neuron, converting only 13% (1 of 7) of CFA neurons vs 88% (7 of 8) of control neurons (p=0.001, Fisher Exact test). Together these results argue that NaV1.7 contributes more and NaV1.8 contributes less to nociceptor excitability after inflammation. (C) At DIV0, PF-71 significantly increased resting membrane potential (T11=-3.530, p=0.005, paired t-test) and rheobase (Z11=2.186, p=0.024, Wilcoxon rank test), and significantly decreased spike height (T11=4.413, p=0.001, paired t-test) in CFA neurons. Further addition of PF-24 significantly changed rheobase (Z9=2.176, p=0.023, Wilcoxon rank test) and spike height (T9=3.237, p=0.01, paired t-test) but did not affect resting membrane potential (T9=1.049, p=0.321, paired t-test).
+
+
+
Results above confirm that NaV1.7 takes on greater responsibility for nociceptor excitability after inflammation, which in turn predicts that PF-71 should reduce pain after inflammation but not under control conditions. As predicted, PF-71 significantly reduced thermal (Fig. 8A) and tactile (Fig. 8B) sensitivity in CFA-inflamed mice without having any effect in control mice. Consistent with this, epigenetic repression of NaV1.7 prevents/reverses hypersensitivity in inflamed and neuropathic mice without causing hyposensitivity in naïve mice (64). This is unlike genetic deletion of NaV1.7, which reduces thermal and tactile sensitivity in naïve mice (27), and with loss-of-function mutations in NaV1.7 that abolish pain in humans (17). These inconsistencies rekindle concerns whether NaV1.7 mutations, unlike pharmacological interventions, affect pain through mechanisms other than modulation of nociceptor excitability. Pharmacological reversal of hypersensitivity in chronic pain conditions (when NaV1.7 is pathologically upregulated) without reducing normal nociceptive pain is clinically desirable, but this hinges on nociceptor hyperexcitability being NaV1.7-dependent, which may be true of some but not all chronic pain conditions, or in only a subset of patients (65).
+
+
+
+DISCUSSION
+
Our results show that nociceptors can achieve similar excitability using different NaV channels. Whereas repetitive spiking depends on NaV1.8 shortly after dissociation (Fig. 2) and presumably under normal conditions in vivo, responsibility shifts to NaV1.7 and NaV1.3 after a few days in vitro (Fig. 3). This is due to translationally regulated changes in NaV expression (Fig. 6). Inflammation causes a similar shift in vivo (Fig. 7). Importantly, acutely inhibiting a particular NaV is consequential (analgesic) only if that subtype is responsible for nociceptor excitability (Fig. 8). This may explain why NaV1.7-selective drugs have not performed well in clinical trials (see Introduction) – because NaV1.7 is not always necessary for nociceptor excitability depending on the expression level of NaV1.7 and other NaV subtypes. Faster processes like channel inactivation also affect their relative contribution (Fig. 5). These observations demonstrate the variable contribution of different NaV subtypes to nociceptor excitability. When unaccounted for, such variability can lead to inconsistencies at the root of poor reproducibility and translatability.
+
+
+
Inflammation-induced change in Na<sub>V</sub> subtype contribution impacts analgesic efficacy of PF-71.
(A) CFA significantly increased thermal sensitivity (F5,65=19.556, p<0.001, two-way RM ANOVA). PF-71 significantly decreased thermal sensitivity in mice injected three days prior with CFA (T8=-7.296, p<0.001; paired t-test) but had no effect in naïve mice (T5=-0.141, p=0.894). (B) CFA significantly increased mechanical sensitivity (F4,52=16.786, p<0.001). PF-71 significantly decreased tactile sensitivity in mice injected three days prior with CFA (T8=-4.341, p=0.002) but had no effect in naive mice (T5=1.000, p=0.363). Insets in both panels show values for each animal before and 2 hours after PF-71 injection. *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc tests.
+
+
+
Contrary to the view that certain ion channels are uniquely responsible for certain aspects of neuronal function, neurons use diverse ion channel combinations to achieve similar function (66, 67). This degeneracy is crucial for enabling excitability and other aspects of neuron physiology to be homeostatically regulated by adjusting ion channels in response to perturbations (68–70). Degeneracy also enables pathological changes in different ion channels to produce equivalent hyperexcitability (71). This is important insofar as similar excitability may belie differences in the underlying ion channels – differences that may render a neuron susceptible or impervious to a drug depending on the functional necessity of the targeted ion channel in that neuron. This is precisely what our data demonstrate in nociceptors. Similar observations have been made in substantia nigra neurons, whose pacemaker activity can be mediated by NaV channels or by voltage-gated calcium channels, meaning TTX may or may not block their spiking (71, 72). Similar interchangeability is evident for the burst firing of Purkinje neurons (73).
+
Degeneracy also exists at the circuit level (74, 75), where it allows differences in the intrinsic excitability of component neurons to be offset (and effectively hidden) by differences in synaptic weights (76). Relevant for pain processing, the spinal dorsal horn circuit can achieve similar output using different synaptic weight combinations (77); specific neuron types may have a greater or lesser impact on circuit function depending on those weights. In effect, degeneracy introduces contingencies. The role of any ion channel in a neuron (or any neuron in a circuit) depends on the other ion channels in that neuron (or the synaptic connections with other neurons in the circuit). Because of such contingencies, a drug may engage its target without producing the intended cellular, circuit or clinical effect. Indeed, different combinations of GABAA receptor activation and chloride driving force can produce equivalent synaptic inhibition (78), but when inhibition is incompensably compromised, the underlying cause necessitates different interventions (79). By this logic, if upregulation of NaV1.7 is responsible for nociceptor hyperexcitability after nerve injury or inflammation, NaV1.7 is an ideal target since “normal” neurons (not reliant on NaV1.7) would be spared the effects of a NaV1.7-selective drug, but the long-term efficacy of such a drug hinges on hyperexcitability remaining NaV1.7-dependent, which cannot be assumed (10). Furthermore, if myelinated afferents (which express minimal NaV1.7 (80)) are responsible for mechanical allodynia under neuropathic conditions (81–84), then NaV1.7-selective drugs should not be expected to alleviate that symptom, which evidently they do not (26), at least not through a direct mechanism. Indeed, ablating nociceptors abolishes acute and inflammatory pain but not neuropathic pain (23, 85). Pathological pain being mediated by more than one afferent type is another example of circuit-level degeneracy.
+
To be interchangeable, NaV subtypes must functionally overlap (86, 87). Indeed, NaV1.8 and NaV1.7 are similar but not identical in their gating properties; for example, their voltage-dependencies partially overlap but the activation curve for NaV1.8 is right-shifted compared to NaV1.7 (88). Consequently, NaV1.7 activates at voltages near threshold whereas NaV1.8 tends to activate at suprathreshold voltages, during initiation and upstroke of the spike, respectively (34, 55). But that separation is not absolute. We found that NaV1.7 contributes to initiation of the first spike in DIV0 neurons, but because it inactivates more readily than NaV1.8, initiation of all subsequent spikes depends on NaV1.8 (see Fig. 5), which activates at perithreshold voltages because voltage threshold is high (depolarized) in the absence of NaV1.7. At DIV4-7, NaV1.7 still inactivates (which causes voltage threshold to rise) but, because of its higher density, continues to produce enough inward current to continue to initiate later spikes. The activation pattern we report for the first spike at DIV0 is consistent with Blair and Bean (89) who quantified the contribution of different NaV channels by recording pharmacologically isolated currents while varying the holding potential according to the spike waveform. Our results go further in showing how responsibilities shift across different spikes within a train (because of differential NaV inactivation) and across conditions (because of changes in NaV expression).
+
With respect to reproducibility, labs testing nociceptors after different times in vitro would be expected to reach contradictory conclusions about the relative importance of a given NaV subtype. Likewise, a testing protocol focusing on single spikes (the equivalent to the first spike in a train) would yield different results from one that considers repetitive spiking. Along the same lines, voltage clamp protocols that deliberately hold membrane potential at unnaturally hyperpolarized voltages to relieve inactivation before stepping up the voltage can give a misleading impression of how much a NaV subtypes contributes under natural conditions (i.e. with natural levels of inactivation). Such discrepancies might be chalked up to irreproducibility if the consequences of those methodological differences are not appreciated, especially if one overlooks how degeneracy allows responsibilities to shift between ion channels. Indeed, the pain literature is replete with apparent inconsistencies. We would argue that most of those studies are correct, but only under limited conditions; failure to identify and report those conditions (contingencies) represents a huge impediment to translation.
+
In summary, our results show that nociceptors can achieve equivalent excitability using different NaV subtypes. The importance of a given subtype can shift on long and short timescales, yielding results that are seemingly inconsistent. By elucidating those shifting responsibilities, our results highlight the degenerate nature of nociceptor excitability and its functional implications. Degeneracy makes it impossible to claim without reservation that a particular NaV subtype is uniquely responsible for pathological pain. Greater appreciation of degeneracy’s implications would prompt better experimental design, more cautious interpretation, and, ultimately, improved translation.
+
+
+MATERIALS AND METHODS
+
+Animals
+
All animal procedures were approved by the Animal Care Committee at The Hospital for Sick Children (protocol #53451) and were conducted in accordance with guidelines from the Canadian Council on Animal Care. We used the Cre-loxP recombinase system to generate mice that express ChR2-eYFP in NaV1.8-expressing neurons. Mice were obtained by crossing homozygous Ai32 mice (B6.Cg-Gt(ROSA)26Sortm32(CAG-COP4*H134R/EYFP)Hze/J) from Jax (#012569), which express ChR2(H134R)-eYFP in the presence of Cre recombinase, with NaV1.8-Cre mice (Tg(Scn10a-cre)1Rkun), which express Cre recombinase in Nav1.8-expressing neurons (kindly provided by Rohini Kuner). These neurons are primarily nociceptive and thermoreceptive (90). The NaV1.8 promoter leads to transgene expression in >90% of neurons expressing markers of nociceptors (21). To ensure that our transgenic mice were typical of wild-type mice with the same background (C57BL/6j), experiments reported in Fig. 1 were repeated in both genotypes for comparison. There was no effect of genotype on rheobase, spike height, input resistance, or spiking pattern, nor was there any significant interaction between genotype and effects of TTX except for spike height at DIV4-7, where TTX had a marginally larger effect in wild-type mice (Two-way ANOVA, F1,54=4.968, p=0.03, see source data file); therefore, we pooled the data for Fig.1. Having verified that our foundational observations held across different genotypes, we used transgenic mice for all subsequent experiments in order to identify eYFP-expressing nociceptors for patching, collection, or imaging.
+
+
+Dorsal root ganglia neuron cultures
+
All key reagents are listed in Supplementary Table 2. Methods for primary DRG culture have been described previously (91). Briefly, adult mice (> 7 week-old) were anaesthetised with isoflurane and perfused intracardiacally with cold HBBS (without Ca and Mg, LifeTech 14170112) supplemented with (in mM) 15 HEPES, 28 Glucose, 111 sucrose, and pH adjusted with NaOH to 7.3-7.4; osmolarity 319-321. Lumbar dorsal root ganglia (DRGs) were extracted (L2-5, except for CFA-inflamed mice, in which we only took L4), digested with papain (Worthington Biochemical Corp.) and collagenase (Worthington Biochemical Corp.)/dispase II (Sigma), and mechanically dissociated by trituration before being plated onto poly-D lysine-coated coverslips and incubated in Neurobasal™ media (Gibco 21103-049) supplemented with 1 % fetal bovine serum (FBS), B-27™ supplement (Thermo Fisher 17504-044) and 0.5mM L-Glutamine (Gibco 25030-081) for an initial period of 2 hours. After this, media was changed to maintenance media (same as plating media but without FBS) and cells were maintained in a 5% CO2 incubator at 37°C. Media was changed every 3-4 days thereafter. Neurons were recorded at two time points after plating: 2-8 hours (referred to as DIV0) or 4-7 days (referred to as DIV4-7).
+
+
+Electrophysiology
+
Coverslips with cultured neurons were transferred from the incubator to a recording chamber perfused with artificial cerebrospinal fluid containing (in mM): 126 NaCl, 2.5 KCl, 2.0 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, 2 MgCl2, and 10 glucose, bubbled with carbogen (5% CO2:95% O2) at room temperature. Neurons were visualized with gradient contrast optics on a Zeiss AxioExaminer microscope using a 40x, 0.75 NA water immersion objective (N-Achroplan, Zeiss) and IR-1000 Infrared CCD camera (Dage-MTI). YFP expression was visualized by epifluorescence (X-Cite, Excelitas) using a Zeiss filter set (46HE). A long-pass filter (OG590) was positioned in the transmitted light path to avoid activating ChR2 while patching. No optogenetic testing was performed as part of this study. Cells expressing YFP and with a soma diameter <25 µm were targeted for whole cell recording using pipettes (∼5 MΩ resistance) pulled from borosilicate glass (WPI).
+
For current clamp recordings, pipettes were filled with intracellular solution containing (in Mm): 140 K-gluconate, 2 MgCl2, 10 HEPES, 0.2 EGTA, 3.8 Na-ATP and 0.4 Na-GTP with pH adjusted to 7.3 with KOH; osmolarity was ∼300 mOsm. A liquid junction potential correction of 15 mV was applied to all reported voltages. Series resistance was compensated to >70%. Signals were amplified with an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, USA), low-pass filtered at 2 kHz, digitized with a Power1401 A/D device (Cambridge Electric Design, Cambridge, UK), and recorded at 10 kHz using CED software Signal version 6. After the natural resting membrane potential was noted, neurons were adjusted to -70 mV using continuous current injection in current clamp mode. Action potentials (spikes) were evoked using a series of 1-second long depolarizing current injections. Rheobase was defined as the minimal current required to evoke a spike. Neurons were tested with current injections from 1x rheobase to 4x rheobase using increments of 0.5x rheobase. Repetitive spiking neurons were defined as those producing ý3 spikes in response to any stimulus intensity; transient spiking neurons consistently produced ≤2 spikes. Spike threshold was defined as voltage where dV/dt first exceeds 5 mV/ms (92). Spike height was measured from threshold to peak of the action potential. Only neurons with a resting membrane potential below -45 mV, spikes overshooting 0 mV and recordings with <20% change in series resistance were tested and analyzed. For dynamic clamp experiments, the pipette shank was painted with Sylgard (Dow) to reduce pipette capacitance. Virtual NaV1.7 and NaV1.8 conductances were introduced into the cells using CED software Signal v6. Currents were defined using the Hodgkin-Huxley equation, using the same parameter values as in our computational model (see below).
+
For voltage-clamp recordings, the bath solution was adjusted to reduce sodium currents to ensure proper clamping (85). Bath solution contained (in mM): 65 NaCl, 50 choline chloride, 5 KCl, 5 HEPES, 5 MgCl2, 10 glucose, and 0.1 CaCl2, plus 0.1 CdCl2 to block calcium currents, and 20 TEA and 5 4-AP to block potassium currents; pH was adjusted to 7.4 with NaOH. Pipettes where filled with intracellular solution containing (in Mm): 140 CsCl, 10 HEPES, 2 MgCl2, 1 EGTA, 3.8 Na-ATP, 0.4 Na-GTP; pH was adjusted to 7.3 with CsOH. The resulting pipette resistance was ∼3 MΩ. A liquid junction potential correction of 4.8 mV was applied to all command voltages. Sodium currents were recorded during 20 ms-long steps from -85 mV to voltages between -45 and +15 mV. Series resistance was compensated to >80%. Signals were amplified, low-pass filtered at 5 kHz, and digitized as described for current clamp recordings.
Cultured DRG neurons <25 μm and expressing eYFP were identified as described above for patching. Coverslips were perfused with aCSF made with DEPC-treated ddH2O, and identified neurons were collected using a glass pipette filled with intracellular solution also made from DEPC-treated ddH2O (composition otherwise the same as described above for electrophysiology). Approximately 50 neurons were collected at DIV0 and at DIV4-7. Total mRNA was extracted with a PureLink RNA mini kit after digestion of genomic DNA with DNase I (Thermo Fisher Scientific) and the cDNA was synthesized with a SuperScript II first-strand synthesis kit (Thermo Fisher Scientific) according to instructions. RT-qPCR was performed with the cDNA primers of target genes (Supplementary Table 3), and the PowerUp SYBR® Green master mix (Thermo Fisher Scientific) in the QuantStudio-3 real-time PCR system. The primers were designed with IDT and spanned at least one exon longer than 1000 bp in order to exclude contamination from genomic RNA. Non-RT mRNA was also used as a negative control to exclude contamination from genomic RNA. All target genes were performed in triplicate for each sample and the experiments were repeated at least 3 times. NaV1.7 and NaV1.8 transcript levels were analyzed using the 2-ΔΔCT method and compared with the housekeeping gene HPRT.
+
+
+Immunocytochemistry
+
Cultured DRG neurons were treated with 4% paraformaldehyde for 10 minutes, rinsed 3x with cold PBS, and permeabilized with 0.1% Triton X-100 in PBS. After another 3x rinse with PBS, neurons were treated with 10% normal goat serum for 30 min followed with rabbit primary NaV1.7 antibody (1:200, ASC-008, Alomone) or NaV1.8 antibody (1:200, ASC-028, Alomone) in PBS with 0.1% Tritween-20 and 1% BSA for 1 h. For some of the coverslips, primary antibodies were replaced with control peptides (ASC008AG1040 for NaV1.7 and ASC016AG0640 for NaV1.8) provided by Alomone as negative controls. Following 3x rinse in PBS, neurons were incubated in the dark with goat anti-rabbit secondary antibody Alexa Fluor-647 (1:500, Abcam) in PBS containing 1% BSA for 1 h, followed by DAPI staining for 10 min. All incubations were done at room temperature. Finally, coverslips were mounted on slides with mounting media (Abcam, ab128982), imaged with a spinning disk confocal microscope (Quorum Technologies) using the same acquisition setting across all imaging sessions, and analyzed with Volocity software (v6.5.1). Protein levels are measured using fluorescence intensity and expressed relative to each other (e.g. ratios in Fig 6C) or relative to fluorescence intensity for YFP in the same cells. Each condition was tested in a minimum of 3 animals.
+
+
+Behavioral testing
+
Behavioral tests were performed on adult mice (male and female, 8-12 weeks). Mice were acclimated to the testing environment for at least 1 h the day prior to start of experiments. Behavioral testing (von Frey test and Hargreaves test) were then performed for 2-3 consecutive days for baseline and for another 3 days after CFA injection. Behavioral tests were performed at the same time in the morning, at room temperature (21°C) following a one hour acclimation period. Animals were randomly assigned to experimental groups and the experimenter was blind to the drug condition.
+
+CFA injection
+
CFA (Sigma, F5881) was thoroughly dissolved in saline (1:1) by vortexing the mixture. The resulting CFA solution (20 µl) was injected subdermally into the left hind paw under light isoflurane anaesthesia. The injection was performed shortly after the last baseline test, on Day 0.
+
+
+PF-71 administration
+
Injectable PF-71 solution was prepared by first dissolving PF-71 in DMSO to make a 5% stock solution; dissolution was achieved by heating to 37°C and vortexing. On the day of injection, stock solution was dissolved in sunflower oil (5% v/v) by sonicating for 5 min. Freshly prepared final PF-71 solution was injected intraperitoneally (1 g/kg body weight). Behavioral testing was performed 2 hours after injection of PF-71 or vehicle.
+
+
+Von Frey testing
+
Mechanical hyperalgesia was assessed with von Frey filaments (North Coast) using the SUDO method (93) . The average of 3 trials in each animal was used for analysis.
+
+
+Hargreaves testing
+
Thermal hyperalgesia was assessed with the Hargreaves apparatus (Ugo Basile, Italy). Radiant heat was applied to the plantar surface of the left hind paw. Interval between stimulus onset and paw withdrawal was defined as paw withdrawal latency (PWL). A 20 s cut-off prevented damage to the skin if the animal failed to withdraw. The average of 3 trials in each animal was used for analysis.
+
+
+
+Statistical analysis
+
Results are expressed as mean ± SEM when data are normally distributed or otherwise as median and quartiles. Normality was tested using the Kolmogorov-Smirnov test. Analysis was performed with GraphPad Prism (v9) and SigmaPlot (v11). Normally distributed data were compared using t-tests or two-way ANOVAs followed by a Student Newman-Keuls post hoc test. Non-normally distributed data were compared using Mann-Whitney and Wilcoxon signed rank tests. Fisher exact and McNemar test were used for categorical data. Exact significance values and test results are reported throughout figure legends.
+
+
+Computer model
+
Two separate, single compartmental models were built for DIV 0 and DIV4-7. They have the same, seven conductances: (𝑔̄Nav1.3, 𝑔Nav1.7, 𝑔Nav1.8, 𝑔Kdr, 𝑔M, 𝑔AHP and 𝑔Leak. Channel equations and their gating parameters are provided in Supplementary Table 4. Conductance densities at baseline (Supplementary Table 5) were tuned to qualitatively reproduce the changes in NaV channel expressions at DIV 0 and 4-7 indicated by the experiments; changes to other channels were minimized between the two models. The effects of ICA, PF-71 and PF-24 were simulated by adjusting (𝑔̄Nav1.3, 𝑔̄Nav1.7 and 𝑔̄Nav1.8=, respectively, as reported in the figures. Channel noise was added as Ornstein-Uhlenbeck process with 𝜇56782= 0 µA/cm>, 𝜎56782= 0.05 µA/cm>, and 𝜏 = 5 ms. All computer code is available at ModelDB (http://modeldb.yale.edu/267560; password: excitability).
+
CONTRIBUTIONS: YX, SR, SAP designed the research; YX, JY collected data, YX, JY, SR, SAP analyzed data; YX, SR, SAP wrote the manuscript.
+
ACKNOWLEDGMENTS: This work was supported by a Restracomp fellowship to JY and by a Canadian Institutes of Health Research Foundation Grant (FDN167276) to SAP. We thank Rohini Kuner for providing Nav1.8-Cre mice, Jason Jeong and Russell Smith for expert technical assistance with animal care and cell cultures, and Yongqian Wang for advice on qPCR data acquisition.
+
+
+
+
+
+REFERENCES
+CohenSP, VaseL, & HootenWM (2021) Chronic pain: an update on burden, best practices, and new advances. Lancet397():2082–2097.
+FinnerupNB, et al. (2015) Pharmacotherapy for neuropathic pain in adults: a systematic review and meta-analysis. Lancet Neurol14():162–173.
+RosenbergerDC, BlechschmidtV, TimmermanH, WolffA, & TreedeRD (2020) Challenges of neuropathic pain: focus on diabetic neuropathy. J Neural Transm127():589–624.
+HayM, ThomasDW, CraigheadJL, EconomidesC, & RosenthalJ (2014) Clinical development success rates for investigational drugs. Nat Biotechnol32():40–51.
+WoolfCJ (2010) Overcoming obstacles to developing new analgesics. Nat Med16():1241–1247.
+MogilJS (2009) Animal models of pain: progress and challenges. Nat Rev Neurosci10():283–294.
+TanejaA, Di IorioVL, DanhofM, & Della PasquaO (2012) Translation of drug effects from experimental models of neuropathic pain and analgesia to humans. Drug Discov Today17():837–849.
+MaoJ (2012) Current challenges in translational pain research. Trends Pharmacol Sci33():568–573.
+EdelmanGM & GallyJA (2001) Degeneracy and complexity in biological systems. Proc Natl Acad Sci U S A98():13763–13768.
+RattéS & PrescottSA (2016) Afferent hyperexcitability in neuropathic pain and the inconvenient truth about its degeneracy. Curr Opin Neurobiol36:31–37.
+GoldMS & GebhartGF (2010) Nociceptor sensitization in pain pathogenesis. Nat Med16():1248–1257.
+HaroutounianS, et al. (2014) Primary afferent input critical for maintaining spontaneous pain in peripheral neuropathy. Pain155():1272–1279.
+YatzivSL & DevorM (2019) Suppression of neuropathic pain by selective silencing of dorsal root ganglion ectopia using nonblocking concentrations of lidocaine. Pain160():2105–2114.
+AllesSRA & SmithPA (2021) Peripheral voltage-gated cation channels in neuropathic pain and their potential as therapeutic targets. Front Pain Res2:750583.
+BeanBP (2007) The action potential in mammalian central neurons. Nat Rev Neurosci. 8():451–465.
+WaxmanSG & ZamponiGW (2014) Regulating excitability of peripheral afferents: emerging ion channel targets. Nat Neurosci17():153–163.
+CoxJJ, et al. (2006) An SCN9A channelopathy causes congenital inability to experience pain. Nature444():894–898.
+FertlemanCR, et al. (2006) SCN9A mutations in paroxysmal extreme pain disorder: allelic variants underlie distinct channel defects and phenotypes. Neuron52():767–774.
+YangY, et al. (2004) Mutations in SCN9A, encoding a sodium channel alpha subunit, in patients with primary erythermalgia. J Med Genet41():171–174.
+Dib-HajjSD, YangY, BlackJA, & WaxmanSG (2013) The NaV1.7 sodium channel: from molecule to man. Nat Rev Neurosci14():49–62.
+NassarMA, et al. (2004) Nociceptor-specific gene deletion reveals a major role for NaV1.7 (PN1) in acute and inflammatory pain. Proc Natl Acad Sci U S A101():12706–12711.
+NassarMA, LevatoA, StirlingLC, & WoodJN (2005) Neuropathic pain develops normally in mice lacking both NaV1.7 and NaV1.8. Mol Pain1():24.
+MinettMS, et al. (2014) Pain without nociceptors? NaV1.7-independent pain mechanisms. Cell Rep6():301–312.
+GrubinskaB, et al. (2019) Rat NaV1.7 loss-of-function genetic model: Deficient nociceptive and neuropathic pain behavior with retained olfactory function and intra-epidermal nerve fibers. Mol Pain15:1744806919881846.
+MinettMS, et al. (2012) Distinct Nav1.7-dependent pain sensations require different sets of sensory and sympathetic neurons. Nat Commun3:791.
+ShieldsSD, et al. (2018) Insensitivity to pain upon adult-onset deletion of NaV1.7 or its blockade with selective inhibitors. J Neurosci38():10180–10201.
+MacDonaldDI, et al. (2021) A central mechanism of analgesia in mice and humans lacking the sodium channel NaV1.7. Neuron109():1497–1512.e1496.
+MinettMS, et al. (2015) Endogenous opioids contribute to insensitivity to pain in humans and mice lacking sodium channel NaV1.7. Nat commun6:8967.
+DehenH, WillerJC, PrierS, BoureauF, & CambierJ (1978) Congenital insensitivity to pain and the “morphine-like” analgesic system. Pain5():351–358.
+EmeryEC, LuizAP, & WoodJN (2016) NaV1.7 and other voltage-gated sodium channels as drug targets for pain relief. Expert Opin Ther Targets20():975–983.
+VetterI, et al. (2017) NaV1.7 as a pain target - From gene to pharmacology. Pharmacol Ther172:73–100.
+YangY, MisMA, EstacionM, Dib-HajjSD, & WaxmanSG (2018) NaV1.7 as a pharmacogenomic target for pain: Moving toward precision medicine. Trends Pharmacol Sci39():258–275.
+KushnarevM, PirvulescuIP, CandidoKD, & KnezevicNN (2020) Neuropathic pain: preclinical and early clinical progress with voltage-gated sodium channel blockers. Expert Opin Investig Drugs29():259–271.
+AlsaloumM, HigerdGP, EffraimPR, & WaxmanSG (2020) Status of peripheral sodium channel blockers for non-addictive pain treatment. Nat Rev Neurol16():689–705.
+EaglesDA, ChowCY, & KingGF (2022) Fifteen years of NaV 1.7 channels as an analgesic target: Why has excellent in vitro pharmacology not translated into in vivo analgesic efficacy?Br J Pharmacol179():3592–3611.
+KitanoY & ShinozukaT (2022) Inhibitio of NaV1.7: the possibility of ideal analgesics. RSC Med Chem13():895–920.
+MulcahyJV, et al. (2019) Challenges and Opportunities for Therapeutics Targeting the Voltage-Gated Sodium Channel Isoform NaV1.7. J Med Chem62():8695–8710.
+BankarG, et al. (2018) Selective NaV1.7 antagonists with long residence time show improved efficacy against inflammatory and neuropathic pain. Cell Rep24():3133–3145.
+KingwellK (2019) NaV1.7 withholds its pain potential. Nat Rev Drug Discov18:321–323.
+AlexandrouAJ, et al. (2016) Subtype-selective small molecule inhibitors reveal a fundamental role for NaV1.7 in nociceptor electrogenesis, axonal conduction and presynaptic release. PLoS One11():e0152405.
+McDermottLA, et al. (2019) Defining the functional role of NaV1.7 in human nociception. Neuron101():905–919.e908.
+RothenbergME, et al. (2019) Safety, tolerability, and pharmacokinetics of GDC-0276, a novel NaV1.7 Inhibitor, in a first-in-human, single- and multiple-dose study in healthy volunteers. Clin Drug Investig39():873–887.
+FlakeNM, LancasterE, WeinreichD, & GoldMS (2004) Absence of an association between axotomy-induced changes in sodium currents and excitability in DRG neurons from the adult rat. Pain109():471–480.
+ZhangJM, DonnellyDF, SongXJ, & LaMotteRH (1997) Axotomy increases the excitability of dorsal root ganglion cells with unmyelinated axons. J Neurophysiol. 78():2790–2794.
+AmirR, MichaelisM, & DevorM (1999) Membrane potential oscillations in dorsal root ganglion neurons: role in normal electrogenesis and neuropathic pain. J Neurosci. 19():8589–8596.
+CaffreyJM, EngDL, BlackJA, WaxmanSG, & KocsisJD (1992) Three types of sodium channels in adult rat dorsal root ganglion neurons. Brain Res592():283–297.
+RenganathanM, CumminsTR, & WaxmanSG (2001) Contribution of NaV1.8 sodium channels to action potential electrogenesis in DRG neurons. J Neurophysiol86():629–640.
+RushAM, CumminsTR, & WaxmanSG (2007) Multiple sodium channels and their roles in electrogenesis within dorsal root ganglion neurons. J Physiol579():1–14.
+ZhangJM, SongXJ, & LaMotteRH (1999) Enhanced excitability of sensory neurons in rats with cutaneous hyperalgesia produced by chronic compression of the dorsal root ganglion. J Neurophysiol82():3359–3366.
+PayneCE, et al. (2015) A novel selective and orally bioavailable Nav 1.8 channel blocker, PF-01247324, attenuates nociception and sensory neuron excitability. Br J Pharmacol172():2654–2670.
+VijayaragavanK, O’LearyME, & ChahineM (2001) Gating properties of NaV1.7 and NaV1.8 peripheral nerve sodium channels. J Neurosci21():7909–7918.
+TheileJW, FullerMD, & ChapmanML (2016) The selective NaV1.7 inhibitor, PF-05089771, interacts equivalently with fast and slow inactivated NaV1.7 channels. Mol Pharmacol90():540–548.
+McCormackK, et al. (2013) Voltage sensor interaction site for selective small molecule inhibitors of voltage-gated sodium channels. Proc Natl Acad Sci U S A110():E2724–2732.
+StregePR, et al. (2017) Sodium channel NaV1.3 is important for enterochromaffin cell excitability and serotonin release. Sci Rep7():15650.
+BennettDL, ClarkAJ, HuangJ, WaxmanSG, & Dib-HajjSD (2019) The role of voltage-gated sodium channels in pain signaling. Physiol Rev99():1079–1151.
+PrescottSA, De KoninckY, & SejnowskiTJ (2008) Biophysical basis for three distinct dynamical mechanisms of action potential initiation. PLoS Comput.Biol. 4():e1000198.
+BertaT, et al. (2008) Transcriptional and functional profiles of voltage-gated Na+ channels in injured and non-injured DRG neurons in the SNI model of neuropathic pain. Mol Cell Neurosci37():196–208.
+DustrudeET, WilsonSM, JuW, XiaoY, & KhannaR (2013) CRMP2 protein SUMOylation modulates NaV1.7 channel trafficking. J Biol Chem288():24316–24331.
+YamaneM, et al. (2017) A functional coupling between CRMP1 and NaV1.7 for retrograde propagation of Semaphorin3A signaling. J Cell Sci130():1393–1403.
+GouldHJ, 3rd, EnglandJD, LiuZP, & LevinsonSR (1998) Rapid sodium channel augmentation in response to inflammation induced by complete Freund’s adjuvant. Brain Res802():69–74.
+BlackJA, LiuS, TanakaM, CumminsTR, & WaxmanSG (2004) Changes in the expression of tetrodotoxin-sensitive sodium channels within dorsal root ganglia neurons in inflammatory pain. Pain108():237–247.
+LiangL, FanL, TaoB, YasterM, & TaoYX (2013) Protein kinase B/Akt is required for complete Freund’s adjuvant-induced upregulation of NaV1.7 and NaV1.8 in primary sensory neurons. J Pain14():638–647.
+AkinEJ, et al. (2019) Building sensory axons: Delivery and distribution of NaV1.7 channels and effects of inflammatory mediators. Sci Adv5():eaax4755.
+MorenoAM, et al. (2021) Long-lasting analgesia via targeted in situ repression of NaV1.7 in mice. Sci Transl Med13(584).
+BaronR & DickensonAH (2014) Neuropathic pain: precise sensory profiling improves treatment and calls for back-translation. Pain155():2215–2217.
+MarderE & GoaillardJM (2006) Variability, compensation and homeostasis in neuron and network function. Nat Rev Neurosci7():563–574.
+O’LearyT, WilliamsAH, FranciA, & MarderE (2014) Cell types, network homeostasis, and pathological compensation from a biologically plausible ion channel expression model. Neuron82():809–821.
+DrionG, O’LearyT, & MarderE (2015) Ion channel degeneracy enables robust and tunable neuronal firing rates. Proc Natl Acad Sci U S A112():E5361–5370.
+MishraP & NarayananR (2022) Conjunctive changes in multiple ion channels mediate activity-dependent intrinsic plasticity in hippocampal granule cells. iScience25():103922.
+YangJ, ShakilH, RattéS, & PrescottSA (2022) Minimal requirements for a neuron to coregulate many properties and the implications for ion channel correlations and robustness. Elife11:e72875.
+DrionG, MassotteL, SepulchreR, & SeutinV (2011) How modeling can reconcile apparently discrepant experimental results: the case of pacemaking in dopaminergic neurons. PLoS Comput Biol7():e1002050.
+PuopoloM, RaviolaE, & BeanBP (2007) Roles of subthreshold calcium current and sodium current in spontaneous firing of mouse midbrain dopamine neurons. J Neurosci27():645–656.
+SwensenAM & BeanBP (2005) Robustness of burst firing in dissociated purkinje neurons with acute or long-term reductions in sodium conductance. J Neurosci25():3509–3520.
+KnoxAT, GlauserT, TenneyJ, LyttonWW, & HollandK (2018) Modeling pathogenesis and treatment response in childhood absence epilepsy. Epilepsia59():135–145.
+PrinzAA, BucherD, & MarderE (2004) Similar network activity from disparate circuit parameters. Nat Neurosci7():1345–1352.
+GrashowR, BrookingsT, & MarderE (2010) Compensation for variable intrinsic neuronal excitability by circuit-synaptic interactions. J Neurosci30():9145–9156.
+MedlockL, et al. (2022) Multiscale computer model of the spinal dorsal horn reveals changes in network processing associated with chronic pain. J Neurosci42():3133–3149.
+PrescottSA, SejnowskiTJ, & De KoninckY (2006) Reduction of anion reversal potential subverts the inhibitory control of firing rate in spinal lamina I neurons: towards a biophysical basis for neuropathic pain. Mol.Pain2.
+LeeKY & PrescottSA (2015) Chloride dysregulation and inhibitory receptor blockade yield equivalent disinhibition of spinal neurons yet are differentially reversed by carbonic anhydrase blockade. Pain156():2431–2437.
+DjouhriL, et al. (2003) Sensory and electrophysiological properties of guinea-pig sensory neurones expressing NaV1.7 (PN1) Na+ channel alpha subunit protein. J Physiol546():565–576.
+CampbellJN, RajaSN, MeyerRA, & MackinnonSE (1988) Myelinated afferents signal the hyperalgesia associated with nerve injury. Pain32():89–94.
+KoltzenburgM, LundbergLE, & TorebjorkHE (1992) Dynamic and static components of mechanical hyperalgesia in human hairy skin. Pain51():207–219.
+LiuCN, et al. (2000) Tactile allodynia in the absence of C-fiber activation: altered firing properties of DRG neurons following spinal nerve injury. Pain85():503–521.
+LiuX, EschenfelderS, BlenkKH, JänigW, & HäblerH (2000) Spontaneous activity of axotomized afferent neurons after L5 spinal nerve injury in rats. Pain84():309–318.
+AbrahamsenB, et al. (2008) The cell and molecular basis of mechanical, cold, and inflammatory pain. Science321():702–705.
+GoaillardJM & MarderE (2021) Ion channel degeneracy, variability, and covariation in neuron and circuit resilience. Annu Rev Neurosci44:335–357.
+YangJ & PrescottSA (2023) Homeostatic regulation of neuronal function: importance of degeneracy and pleiotropy. Front Cell Neurosci17:1184563.
+SchildJH & KunzeDL (1997) Experimental and modeling study of Na+ current heterogeneity in rat nodose neurons and its impact on neuronal discharge. J Neurophysiol78():3198–3209.
+BlairNT & BeanBP (2002) Roles of tetrodotoxin (TTX)-sensitive Na+ current, TTX-resistant Na+ current, and Ca2+ current in the action potentials of nociceptive sensory neurons. J Neurosci22():10277–10290.
+AgarwalN, OffermannsS, & KunerR (2004) Conditional gene deletion in primary nociceptive neurons of trigeminal ganglia and dorsal root ganglia. Genesis38():122–129.
+MalinSA, DavisBM, & MolliverDC (2007) Production of dissociated sensory neuron cultures and considerations for their use in studying neuronal function and plasticity. Nat Protoc2():152–160.
+DavidsonS, et al. (2014) Human sensory neurons: Membrane properties and sensitization by inflammatory mediators. Pain155():1861–1870.
+BoninRP, BoriesC, & De KoninckY (2014) A simplified up-down method (SUDO) for measuring mechanical nociception in rodents using von Frey filaments. Mol Pain10:26.
+
+
+
+
+
Inhibiting Na<sub>V</sub>1.8 at DIV0 with A-803467 had the same effect as PF-24.
(A) Sample voltage clamp recording at DIV0 before and after A-803467 (1 µM). (B) Peak current was significantly reduced by A-803467 (F1,84=9.935, p=0.016, two-way RM ANOVA, n=8). *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc test in B.
+
+
+
+
+
Inhibiting Na<sub>V</sub>1.8 with PF-24 at DIV4-7 had negligible effects.
(A) Inhibiting Nav1.8 with PF-24 (1 µM) did not affect spiking pattern (χ2 =0.00, p=1.00, McNemar test) and modestly reduced firing rate (F1,54=9.745, p= 0.012, two-way RM ANOVA, n=10) in DIV4-7 neurons. (B) PF-24 did not affect rheobase (Z12=0.420, p=0.685, Wilcoxon Rank test) but did reduce spike height (T12=2.939, p=0.012, paired-t-test). *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc tests in A.
+
+
+
+
+
Inhibiting Na<sub>V</sub>1.7 at DIV0 had negligible effects.
(A) Inhibiting NaV1.7 with PF-71 (30 nM) did not alter spiking pattern (χ2=0.00, p=1.00, McNemar test) or reduce firing rate (F1,30=5.805, p=0.061, two-way RM ANOVA, n=6) in DIV0 neurons; in fact, firing rate was slightly increased. (D-E) PF-71 did not affect rheobase (Z9=0.677, p=0.578, Wilcoxon rank test) but did reduce spike height (T9=3.759, p=0.004, paired-t-test).
+
+
+
+
+
Na<sub>V</sub>1.7 and Na<sub>V</sub>1.3 currents can compensate for each other.
(A) In the DIV4-7 model, reducing 𝑔̄Nav1.7 to 30% of its normal value (35®10 mS/cm2) can be compensated for by increasing 𝑔̄Nav1.3 to 229% of its normal value (0.35®0.8 mS/cm2) to maintain repetitive spiking. (B) Conversely, reducing 𝑔̄Nav1.3 to 10% of its normal value (0.35®0.035 mS/cm2) can be compensated for by increasing 𝑔̅Nav1.7 to 171% of its normal value (35®60 mS/cm2) and maintain repetitive spiking.
+
+
+
+
+
Virtual conductances restored repetitive spiking after pharmacological inhibition of the corresponding native conductance had converted the neuron to transient spiking.
(A) Sample response at DIV0 showing that a virtual NaV1.8 conductance applied with dynamic clamp restored repetitive spiking after inhibiting native NaV1.8 channels with PF-24. This restoration was repeated in 3 of 3 neurons tested. (B) Sample recording at DIV4-7 showing that a virtual NaV1.7 conductance restored repetitive spiking after inhibiting native NaV1.7 channels with PF-71. This restoration was repeated in 4 of 4 neurons tested.
+
+
+
+
+
Channel inactivation affects Na<sub>V</sub> subtype contribution on short timescale.
(A) In the DIV0 model, NaV1.7 contributed to the first spike but its inactivation meant that all subsequent spikes relied exclusively on NaV1.8. (B) In the DIV4-7 model, despite some inactivation of NaV1.3 (maroon) and NaV1.7 (orange), the remaining current was still large enough (because of the higher gmax of those two subtypes) to produce inward current sufficient to support repetitive spiking despite the low gmax of NaV1.8 in the DIV4-7 model.
+
+
+
+
+
Effect of TTX on voltage threshold in DIV0 neurons.
Despite TTX having negligible effects in DIV0 neurons according to our initial analysis (see Fig. 1), simulation results in Fig. 5A,B predicted that the first spike was nonetheless initiated by NaV1.7. By extension, this predicted that TTX should cause a depolarizing shift in voltage threshold for the first spike. Analysis of the experimental data confirmed this to be true, with threshold (mean±SEM) increasing from -33.7±1.4 mV at baseline to -28.3±1.4 mV after TTX (T24=-3.19, p=0.004, paired t-test). Confirmation of this unexpected prediction helps further validate our model neuron.
+
+
+
+
Model data before and after channel “inhibition”
+
+
+
+
Reagents
+
+
+
+
Primers
+
+
+
+
Model equations
+
+
+
+
+
Conductance densities at baseline for DIV 0 and 4-7 models
+
+
+
+
+
diff --git a/test/2023.06.26.546606/various-fixes/2023.06.26.546606.xml b/test/2023.06.26.546606/various-fixes/2023.06.26.546606.xml
new file mode 100644
index 00000000..5a335ac4
--- /dev/null
+++ b/test/2023.06.26.546606/various-fixes/2023.06.26.546606.xml
@@ -0,0 +1,548 @@
+
+
+
+
+
+BIORXIV
+
+bioRxiv
+bioRxiv
+
+
+Cold Spring Harbor Laboratory
+
+
+
+10.1101/2023.06.26.546606
+1.1
+
+
+Regular Article
+
+
+New Results
+
+
+Cancer Biology
+
+
+
+Metabolic reprogramming of cancer cells by JMJD6-mediated pre-mRNA splicing is associated with therapeutic response to splicing inhibitor
+
+
+
+JablonowskiCarolyn
+1
+#
+
+
+QuarniWaise
+1
+#
+
+
+SinghShivendra
+1
+
+
+TanHaiyan
+2
+
+
+BostanthirigeDhanushka Hewa
+1
+
+
+JinHongjian
+3
+
+
+FangJie
+1
+
+
+ChangTi-Cheng
+2
+
+
+FinkelsteinDavid
+3
+
+
+ChoJi-Hoon
+2
+
+
+HuDongli
+1
+
+
+PagalaVishwajeeth
+2
+
+
+SakuradaSadie Miki
+4
+
+
+Pruett-MillerShondra M.
+4
+
+
+WangRuoning
+5
+
+
+MurphyAndrew
+1
+
+
+FreemanKevin
+6
+
+
+PengJunmin
+2
+
+
+DavidoffAndrew M
+1
+7
+8
+
+
+http://orcid.org/0000-0002-1678-5864
+WuGang
+3
+
+
+YangJun
+1
+7
+8
+9
+
+Department of Surgery, St Jude Children’s Research Hospital, Memphis, TN38105, USA
+Center for Proteomics and Metabolomics, Department of Structural Biology, Department of Developmental Neurobiology, St Jude Children’s Research Hospital, Memphis, TN38105, USA
+Center for Applied Bioinformatics, St Jude Children’s Research Hospital, Memphis, TN38105, USA
+Department of Cell and Molecular Biology, St. Jude Children’s Research Hospital, Memphis, TN 38105, USA
+Center for Childhood Cancer and Blood Disease, Abigail Wexner Research Institute, Nationwide Children’s Hospital, 700 Children’s Drive, Columbus, OH 43205, USA
+Genetics, Genomics & Informatics, The University of Tennessee Health Science Center (UTHSC), Memphis, TN 38103, USA
+St Jude Graduate School of Biomedical Sciences, St Jude Children’s Research Hospital, TN 38105
+Department of Pathology, College of Medicine, The University of Tennessee Health Science Center, 930 Madison Ave, Suite 500, Memphis, TN 38163
+
+
+
equal contribution
+Correspondence: Jun.Yang2@stjude.org
+
The authors declare no potential conflicts of interest
Dysregulated pre-mRNA splicing and metabolism are two hallmarks of MYC-driven cancers. Pharmacological inhibition of both processes has been extensively investigated as potential therapeutic avenues in preclinical and clinical studies. However, how pre-mRNA splicing and metabolism are orchestrated in response to oncogenic stress and therapies is poorly understood. Here, we demonstrate that JMJD6 acts as a hub connecting splicing and metabolism in MYC-driven neuroblastoma. JMJD6 cooperates with MYC in cellular transformation by physically interacting with RNA binding proteins involved in pre-mRNA splicing and protein homeostasis. Notably, JMJD6 controls the alternative splicing of two isoforms of glutaminase (GLS), namely kidney-type glutaminase (KGA) and glutaminase C (GAC), which are rate-limiting enzymes of glutaminolysis in the central carbon metabolism in neuroblastoma. Further, we show that JMJD6 is correlated with the anti-cancer activity of indisulam, a “molecular glue” that degrades splicing factor RBM39, which complexes with JMJD6. The indisulam-mediated cancer cell killing is at least partly dependent on the glutamine-related metabolic pathway mediated by JMJD6. Our findings reveal a cancer-promoting metabolic program is coupled with alternative pre-mRNA splicing through JMJD6, providing a rationale to target JMJD6 as a therapeutic avenue for treating MYC-driven cancers.
+
+
+
+
+
+
+
+Competing Interest Statement
The authors have declared no competing interest.
+
+
+
+
+Introduction
+
Metabolic reprogramming is a hallmark of cancer1-3 which allows rapidly proliferating tumor cells to acquire nutrients to meet their bioenergetic, biosynthetic, and redox demands4. One of the primary driving forces in reprograming cancer cell metabolism is the deregulated MYC family proto-oncogenes (C-MYC, MYCN, and MYCL)5, which are known to encode master transcriptional factors that regulate metabolic gene expression. MYC coordinates nutrient acquisition to produce ATP and key cellular building blocks that increase cell mass and promote DNA replication and cell division6. The increase in total RNA and protein synthesis by overactive MYC signaling leads to dysregulation of macromolecular processing machineries including the spliceosome7, and consequently pre-mRNA splicing8-10, another hallmark of MYC-driven cancers7,9-12, for the purpose of cellular stress adaptation. MYCN amplification is one of the most important biological features of high-risk neuroblastoma13. Transgenic mouse and zebrafish models have demonstrated that MYCN is a neuroblastoma driver14,15. In tumors without MYCN amplification, C-MYC is overexpressed, further indicating that neuroblastoma is a MYC-driven cancer. The metabolic dependency of neuroblastoma has been widely studied by us and others16-23. A larger number of splicing changes have also been noticed in high-stage neuroblastomas24-26. Splicing alterations lead to a spliceosomal vulnerability that provides a new opportunity to develop transformative therapies by disrupting aberrant pre-mRNA splicing7,9-12. We and others have shown that targeting the splicing factor RBM39 by indisulam, a “molecular glue” that bridges RBM39 to E3 ubiquitin ligase DCAF15 for proteasomal degradation, achieved an exceptional anti-tumor activity in neuroblastoma models27,28. Disruption of spliceosome by Pladienolide B also resulted in significant anti-tumor effect in neuroblastoma models26 However, how the dysregulated pre-mRNA splicing machinery and metabolism are orchestrated in MYC-driven neuroblastoma has not been well elucidated. Whether metabolism modulates the anti-cancer effect of splicing inhibition remains to be answered.
+
Next-generation sequencing studies have revealed only a few recurrent somatic mutations in neuroblastoma at the time of diagnosis29,30. However, copy number alterations of chromosomal segments such as 17q gain, 1p36 or 11q23 loss frequently occur in high-risk neuroblastoma. While attempts to understand the functions of individual genes in these chromosomal segments have been reported (i.e., BIRC531, PHB32, PPM1D33, TRIM3734 in 17q; ARID1A35, CAMTA136, CASZ137, CHD538-40, KIF1Bβ41-43, miR-34a44, RUNX345 in 1p36), the biological consequences of these genetic events in MYC-driven tumors still remain largely unknown. Gain of 17q is the most frequent genetic event in high-risk neuroblastoma and is associated with MYCN amplification46. In addition, in the transgenic MYCN mouse model of neuroblastoma, the chromosomal locus syntenic to human 17q is partially amplified47, indicating that chromosome 17q is needed for MYC-mediated tumorigenesis.
+
JMJD6 is a JmjC domain–containing nuclear protein with iron- and 2-oxoglutarate–dependent dioxygenase activity48, whose coding gene is located on chromosome 17q25. While the histone arginine demethylase activity of JMJD6 that catalyzes demethylation of H4R3me1/me2 is controversial49, JMJD6 is a lysyl-5-hydroxylase that catalyzes 5-hydroxylation on specific lysine residues of target proteins50. JMJD6 has pleiotropic functions in normal physiology and in cancer51-54. We previously found that JMJD6 is essential for the survival of neuroblastoma cells (including MYCN-amplified and C-MYC– overexpressed cells)55, which was further validated by an independent study56, indicating that neuroblastoma has JMJD6 dependency. However, the exact mechanism of JMJD6 in MYC-driven cancers remains elusive. One study has shown that JMJD6 and BRD4 co-bind at antipause enhancers, regulating promoter-proximal pause release of a large subset of transcription units57. By harnessing a similar mechanism, JMJD6 promotes cell survival of glioblastoma in vivo58. These findings are particularly interesting because BRD4 occupies exceptionally large super-enhancers associated with genes, including C-MYC and MYCN59-61, and the expression of those enhancers can be disrupted by BRD4 inhibitors, which have a potent antitumor effect59-62. Here we show a new mechanism by which JMJD6 promotes tumorigenesis mediated by the MYC oncogene in that JMJD6 interacts with a subset of RNA binding proteins including RBM39 in neuroblastoma cells and regulates the alternative splicing of metabolic genes that are involved in mitochondrial metabolism. “Glutamine addiction” is one key feature of MYC-driven tumors. Glutaminase (GLS) is the enzyme responsible for conversion of glutamine to glutamate in the process of glutaminolysis to feed the tricarboxylic acid (TCA) cycle and has two splice isoforms, GAC (glutaminase C) and KGA (kidney-type glutaminase). We show that JMJD6 controls the alternative splicing of KGA and GAC, and, consequently, impacts the central carbon metabolism in neuroblastoma. Further we show that JMJD6 is correlated with the anti-cancer activity of indisulam, a “molecular glue” that degrades the splicing factor RBM39. The indisulam-mediated cancer cell killing is at least partly dependent on the glutamine-related metabolic pathway mediated by JMJD6. Our findings demonstrate a new mechanism by which JMJD6 coordinates metabolic programs and alternative pre-mRNA splicing, providing a rationale to target JMJD6 as a therapeutic target for MYC-driven cancers.
+
+
+Results
+
+The essential genes for neuroblastoma cell survival on chromosome 17q target pre-mRNA splicing and metabolism
+
An incomplete understanding of the biological consequences of chromosome 17q gain remains a barrier to the understanding of high-risk neuroblastoma. 1132 genes are located on 17q (Supplementary table 1). We surmised that some of the 17q genes are particularly important for neuroblastoma cell survival. Analysis of the cancer dependency genes in neuroblastoma cell lines screened with the Avana sgRNA library63 revealed that 114 were essential to neuroblastoma (mean score <-0.4) (Figure 1a, Supplementary table 2). Protein interaction network analysis followed by functional annotation revealed that proteins encoded by these 114 essential genes formed distinct but interconnected modules including RNA splicing (i.e., SRSF2, DDX5, DDX42, DHX8), mitochondrial metabolism (i.e., NDUFA8, COX11, SLC25A10, SLC35B1), protein homeostasis (i.e., UBE2O, PSMB3, PSMC5), DNA repair (i.e., BRIP1, BRCA1, RAD51C, RAD51D) and transcriptional regulation (i.e., PHF12, CBX1, SMARCE1, MED1), as well as endocytosis (i.e., CHMP6, CTLC, EPN3, HGS, SNF8, VPS25) (Figure 1b). Using these 114 genes as a signature, we found that 81 of them were highly expressed in high-risk neuroblastomas, which were enriched with MYCN amplification (Figure 1c). Correspondingly, neuroblastomas with high expression levels of this gene signature were associated with a poorer event-free and overall survival of patients (Figure 1d, e). These data demonstrate that 17q genes are involved in essential biological processes and highly expressed in high-risk neuroblastomas.
+
+
+
17q contains neuroblastoma dependency genes
+
a. CRISPR score for 17q genes in 10 neuroblastoma cell lines. Score <-0.4 is defined as neuroblastoma dependency genes. Data are derived from Avana sgRNA library screening63.
+
b. STRING protein interaction network showing 17q essential genes with various biological functions.
+
c. Heatmap by K-means clustering analysis showing 17q essential genes are highly expressed in high-risk neuroblastomas based on RNA-seq data (SEQC dataset).
+
d. Kaplan-Meier survival curve showing 17q essential gene signature is correlated with worse event-free survival (SEQC dataset).
+
e. Kaplan-Meier survival curve showing 17q essential gene signature is correlated with worse overall survival (SEQC dataset).
+
+
+
+
+
+JMJD6 is required for neuroblastoma growth
+
JMJD6 was among these 114 essential genes. To understand the role of JMJD6, we examined the genetic features of JMJD6 in neuroblastoma and other types of cancers. Among the genes encoding JmjC-domain containing proteins, JMJD6 was the only one that was frequently amplified in neuroblastoma (Figure 2a). High JMJD6 expression was associated with poor event-free outcome, as shown by Kaplan-Meier analysis (Figure 2b). To examine whether JMJD6 amplification is limited to specific tumor types, we explored genomic data from different cancers using the cBioportal program. JMJD6 was amplified across multiple types of adult cancers such as breast and liver cancer (Supplementary Figure 1a), and correlated with worse relapse-free survival (Supplementary Figure 1b). We further compared the RNA-seq expression of JMJD6 in 2337 samples across over 20 pediatric cancer subtypes and found that JMJD6 showed the highest expression levels in neuroblastoma (Supplementary Figure 1c), suggesting that JMJD6 might be particularly important in neuroblastoma. We validated this hypothesis using shRNA knockdown of JMJD6 in MYCN amplified cells (BE2C, SIMA, KELLY, IMR32) and non-MYCN amplified cells (SK-N-AS and CHLA20). The results showed that loss of JMJD6 greatly reduced the colony numbers in all tested cell lines (Figure 2c), demonstrating that JMJD6 is essential to neuroblastoma cells regardless of MYCN amplification. Neuroblastic tumors comprise a histologic spectrum that ranges from less-differentiated neuroblastoma to well-differentiated ganglioneuroma. The extent of differentiation in the tumor cells is correlated with prognostic significance64. We noticed that the loss of JMJD6 led to neurite outgrowth (Supplementary Figure 1d), a unique structure of neuroblastoma cells differentiating in vitro. This morphologic change suggests that JMJD6 is required to maintain cancer cell stemness. Lastly, we validated that JMJD6 is essential to neuroblastoma growth in MYCN amplified (BE2C) and C-MYC overexpressed (SK-N-AS) xenograft models (Figure 2d, e). Taken together, these data demonstrate that loss of JMJD6 function impedes neuroblastoma cell survival and tumor growth.
+
+
+
JMJD6 is required for neuroblastoma growth and facilitates MYC-mediated cellular transformation.
+
a. Copy number of genes encoding JmjC domain proteins in St Jude neuroblastoma cohort.
+
b. Kaplan-Meier survival curve showing high JMJD6 is correlated with worse event-free survival (SEQC RNA-seq dataset).
+
c. Crystal violet showing the colony staining after JMJD6 siRNA knockdown in neuroblastoma cell lines validated by western blot.
+
d. Xenograft tumor growth of BE2C (right) models with lentiviral JMJD6 shRNA knockdown. P-value calculated by multiple unpaired t-test across each row. n=5 per group.***p<0.001, **p<0.01.
+
e. Xenograft tumor growth of SK-N-AS models with lentiviral JMJD6 shRNA knockdown. P-value calculated by multiple unpaired t-test across each row. ***p<0.001, **p<0.01.
+
f. Western blot validating the expression of retroviral based MYCN and JMJD6 in JoMa1 cells.
+
g. Cell proliferation of JoMa1 cells transduced with indicated constructs, GFP, JMJD6, MYCN, JMJD6+MYCN.
+
h. Colony formation JoMa1 cells transduced with indicated constructs, GFP, JMJD6, MYCN, JMJD6+MYCN. Top panel showing photos taken under light microscope. Bottom panel showing cell colonies stained with crystal violet. P value calculated by multiple unpaired t-test across each row. ***p<0.001, **p<0.01.
+
i. Xenograft tumor growth of JoMa1 cells transduced with indicated constructs, GFP, JMJD6, MYCN, JMJD6+MYCN. n=5 per group. P value calculated by multiple unpaired t-test across each row. ***p<0.001, **p<0.01.
Next, we investigated whether gain of function of JMJD6 could facilitate MYC-mediated oncogenic transformation. To test this, we used a NIH3T3 transformation assay that provides a straightforward method to assess the transforming potential of an oncogene, which may lead to morphological transformation and loss of contact inhibition, a typical feature of cellular transformation. Like the GFP control (Supplementary Figure 2a), we found that NIH3T3 cells with overexpressed JMJD6 stopped proliferation after being confluent (Supplementary Figure 2b), indicating JMJD6 alone is unable to transform NIH3T3 cells. However, overexpression of MYCN induced foci formation with enhanced cell death (Supplementary Figure 2c). Importantly, NIH3T3 cells with overexpressed MYCN and JMJD6 lost contact inhibition, accompanied with morphological change, and formed larger foci (Supplementary Figure 2d), indicating that JMJD6 enhances MYCN activity to transform NIH3T3 cells. Interestingly, MYCN alone also reprogramed metabolism of NIH3T3 cells as shown by the color change of the media, indicative of acidic pH change, which was largely rescued by co-expression of JMJD6 (Supplementary Figure 2e), suggesting that cells with enhanced lactate production by MYCN were directed to oxidative phosphorylation by JMJD6. It is believed that the cell of origin of neuroblastoma is the progeny of neural crest cells. We therefore tested the role of JMJD6 in MYC-mediated transformation using JoMa1, a cell line derived from murine neural crest, by transducing GFP, JMJD6, MYCN and JMJD6/MYCN (Figure 2f). While JMJD6 showed no difference from GFP control in regulating cell proliferation, MYCN slightly enhanced cell proliferation (Figure 2g). However, the combination of JMJD6 and MYCN remarkably increased cell proliferation, mirrored by the colony formation assay which showed JMJD6/MYCN induced rapid growth of colonies with distinct transformation morphology (Figure 2g, h). Implantation of each group into immune-deficient mice led to tumor development of MYCN and JMJD6/MYCN groups (Figure 2i). However, JMJD6/MYCN group tumors appeared to grow faster than the MYCN alone tumors. Taken together, these data indicate that JMJD6 enhances the MYC-mediated transformation, demonstrating the oncogenic role of JMJD6.
+
+
+JMJD6 regulates pathways engaged in pre-mRNA splicing and mitochondrial biogenesis in neuroblastoma
+
We surmised that loss of function of genes in the same functional module/pathway may induce similar effects across different cell lineages, which in turn corroborates the hypothesis that JMJD6 is a player in that signaling pathway. To test this hypothesis, we analyzed the dependency correlation of JMJD6 knockout and other genes (defined as co-dependency genes if they are positively correlated), by using the DepMap data (https://depmap.org) that includes genome-wide knockout in 1027 cell lines of more than 20 cancer types (Supplementary Table 3), followed by pathway enrichment. The data showed that JMJD6 co-dependency genes were significantly and positively correlated with spliceosome/mRNA splicing (i.e., RBM39, SF3B1), ubiquitin-mediated proteolysis and endocytosis and a number of 17q25 genes, (Figure 3a, c), which mirrored the pathway network of 17q essential genes in neuroblastoma (Figure 1b). JMJD6 knockout was negatively correlated with the knockout of genes housed at chromosome 1p, which is frequently deleted in high-risk neuroblastoma, and oxidative phosphorylation as well as protein translation (Figure 3b, d).
+
+
+
JMJD6 regulates pre-mRNA splicing of genes involved in metabolism
+
a. Pathway enrichment for JMJD6 co-dependency genes whose knockout exhibits similar phenotype with JMJD6 knockout based on re-analysis of DepMAP data.
+
b. Pathway enrichment for genes whose knockout exhibits opposite phenotype with JMJD6 knockout based on re-analysis of DepMAP data.
+
c. Chromosomal location enrichment for JMJD6 co-dependency genes whose knockout exhibits similar phenotype with JMJD6 knockout based on re-analysis of DepMAP data.
+
d. Chromosomal location enrichment for genes whose knockout exhibits opposite phenotype with JMJD6 knockout based on re-analysis of DepMAP data.
+
e. Pathways analysis for genes downregulated and upregulated by JMJD6 knockdown commonly shared in SK-NAS and BE2C cells.
+
f. Alternative splicing events altered by JMJD6 knockdown in BE2C and SK-N-AS cells.
+
g. Pathway enrichment for each splicing event commonly shared in BE2C and SK-N-AS cells after JMJD6 knockdown.
+
h. Isoform identification based on splicing events in BE2C and SK-N-AS cells, followed by pathway enrichment for commonly shared alterations in both cell lines.
+
+
+
+
BRD4 is known to regulate MYC expression61,65. Previous studies have shown that JMJD6 and BRD4 interact to regulate gene transcription56,58,66, suggesting that JMJD6 might directly modulate MYC expression. To assess this possibility, we knocked down JMJD6 in neuroblastoma cell lines BE2(C) and SK-N-AS, which express MYCN and C-MYC, respectively, for RNA-seq analysis. The sequencing data were analyzed for differential gene expression (Supplementary Table 4). Interestingly, loss of JMJD6 showed minimal impact on expression of MYCN in BE2C cells or C-MYC in SK-N-AS cells (Supplementary Figure 3a), and western blot analysis did not show alteration of MYCN expression although C-MYC protein was slightly downregulated by loss of JMJD6 (Supplementary Figure 3b). However, BRD4 inhibitors drastically inhibited both MYCN and C-MYC expression in neuroblastoma cells59-61, suggesting that JMJD6 inhibition has a distinct effect from the BRD4 inhibition. Gene set enrichment analysis for pathway engagement for the genes commonly downregulated or upregulated in both cell lines revealed that loss of JMJD6 most significantly repressed the expression of genes involved in pre-mRNA splicing, histones, and cell cycle G1/S checkpoint (Figure 3e), and enhanced the pathways involved in mitochondrial functions and heat shock response (Figure 3e). Interestingly, the genes transcribed from the mitochondrial genome were elevated in both cell lines (Supplementary Figure 3c), suggesting that JMJD6 directly or indirectly regulates the transcription of mitochondrial genome. These data are consistent with the co-dependency pathways of JMJD6 (Figure 3a-d). Depletion of JMJD6 in both cell lines led to the downregulation of MYC signaling pathways (Supplementary Figure 4), although ranked behind the pathways of splicing and metabolism, suggesting that the MYC pathways are not primarily regulated by JMJD6. These data indicate that JMJD6 does not regulate the gene expression of the MYC family of transcription factors but might indirectly regulate the MYC pathway.
+
To verify if JMJD6 regulates pre-mRNA splicing, we analyzed the RNA-seq using two algorithms. The first one is event-based analysis to identify the altered exon splicing (Figure 3f). We found that knockdown of JMJD6 dominantly affects the exon skipping (SE) although other splicing events were also altered, albeit with a much smaller number (Figure 3f). Pathway analysis of common events in both BE2C and SK-N-AS cells demonstrated that genes involved in metabolism and splicing are the most significantly affected by loss of function of JMJD6 (Figure 3g). Using the second algorithm of RNA splicing analysis previously developed (Figure 3h) that allows discovery of new isoforms of genes generated through alternative splicing67, we identified 580 genes in BE2C cells and 1018 genes in SK-N-AS cells undergoing alternative splicing after JMJD6 knockdown, 133 of which were shared by both (Figure 3h, Supplementary Fig. 5, Supplementary Table 5). The alternatively spliced genes were involved in a variety of pathways (Supplementary Table 6). KEGG pathway enrichment analysis of the 133 commonly alternatively spliced genes showed that only metabolic pathway genes were significantly enriched (Figure 3f), most of which are involved in mitochondrial bioenergetics and folate metabolism (Figure 3h). Collectively, these data demonstrate that JMJD6 regulates RNA splicing of genes engaged in mitochondrial metabolism, being one of the key mediators of the 17q locus activity in neuroblastoma.
+
+
+JMJD6 regulates alternative splicing of glutaminolysis gene GLS
+
Overactive MYC signaling leads to altered macromolecular processing machineries in response to an increase in total RNA and protein synthesis7. MYC is also a master regulator of cancer metabolism involved in ribosomal and mitochondrial biogenesis, glucose and glutamine metabolism and lipid synthesis, leading to the acquisition of bioenergetic substrates enabling the cancer cell to grow and proliferate68-70. “Glutamine addiction” is one feature of MYC-driven cancer71. The pre-mRNA splicing altered by JMJD6 knockdown included GLS, the key enzyme of glutaminolysis, prompting us to investigate the GLS splicing mediated by JMJD6 knockdown. GLS is known to catalyze the conversion of glutamine to glutamate, and is alternatively spliced to form two isoforms, GAC and KGA72, with different cellular localization and catalytic capacities73. The GAC isoform is more frequently upregulated in cancer cells than KGA74, and has been shown to be regulated by MYC75-77, leading to a “glutamine addiction” phenotype in MYC-driven tumors71. We found that loss of JMJD6 led to a splicing switch from the GAC isoform (with exons 1-15) to the KGA isoform (with exons 1-14 and 16-19) (Figure 4a), which was further confirmed by isoform-specific RT PCR (Figure 4b). We then investigated the expression of GAC/KGA at the protein levels after JMJD6 knockdown. Western blot showed that the KGA isoform was increased after JMJD6 knockdown in all three tested cell lines, SK-N-AS, BE2C and SIMA (Figure 4c). Then we further validated the JMJD6 effect on GLS isoform expression by using a luciferase reporter that indicates the isoforms of GAC and KGA. Indeed, JMJD6 knockdown significantly increased the expression of the KGA reporter (Figure 4d). RNA-immunoprecipitation showed that JMJD6 bound to GLS RNA (Figure 4e), suggesting that JMJD6 may directly regulate the splicing of GLS. We reasoned that if the regulation of GLS splicing by JMJD6 was a bone fide mechanism, the expression levels of JMJD6 would correlate with the levels of GAC/KGA in tumors. Indeed, JMJD6 was positively correlated with GAC and negatively correlated with KGA in two independent neuroblastoma cohorts (Figure 4f), supporting the hypothesis that JMJD6 is required to maintain the high ratio of GAC/KGA in cancer cells by controlling their alternative splicing. Clinical relevance of GAC and KGA in neuroblastoma was evidenced by the findings that high GAC was associated with a worse event-free survival and high KGA was associated with a better event-free survival (Figure 4g), suggesting that the GAC/KGA ratio may play a role in cancer progression. However, introduction of either GAC or KGA in BE2C cells or SKNAS cells promoted colony formation (Figure 4h, i), indicating that enhanced glutaminolysis by either GAC or KGA overexpression is pro-proliferative by gaining more ATPs.
+
+
+
JMJD6 regulates alternative splicing of glutaminolysis gene, GLS
+
a. Sashimi plot showing the alternative splicing of GLS after JMJD6 knockdown in BE2C and SK-N-AS cells in duplicates. The number indicates the RNA-seq read counts of exon junction.
+
b. Real time PCR assessing the relative expression of GAC and KGA isoforms after JMJD6 knockdown in triplicates.
+
c. Western blot showing the expression of GAC and KGA isoforms in SK-N-AS, BE2C, SIMA after JMJD6 knockdown for 72 hours.
+
d. KGA and GAC specific reporter analysis showing only KGA-driven luciferase activity is significantly upregulated by JMJD6 knockdown.
+
e. RNA-immunoprecipitation showing JMJD6 interaction with GLS RNA. Top panel shows the western blot analysis of Flag tagged JMJD6 in input, immunoprecipitation (IP) and flowthrough (FT)_fractions. Bottom panel shows RT PCR analysis of enrichment of GAC/KGA bound by JMJD6 in IP and FT fractions.
+
f. Spearman correlation analysis of JMJD6 and GAC/KGA expression levels in two neuroblastoma cohorts GSE45547(left) and GSE120572 (right).
+
g. Kaplan-Meier curve showing the association of GAC and KGA expression levels with event-free survival in a cohort of neuroblastoma (GSE45547).
+
h. Western blotting analysis of expression of KGA and GAC in BE2C cells with indicated antibodies.
+
i. Colony formation assay for BE2C cells overexpressing KGA and GAC for 7 days (left =crystal violet staining, right = quantification of cell density). **p<0.001,
+
j. Western blotting analysis of expression of KGA and GAC in SK-N-AS cells with indicated antibodies.
+
k. Colony formation assay for SK-N-AS cells overexpressing KGA and GAC for 7 days (left =crystal violet staining, right = quantification of cell density). **p<0.001,
+
+
+
+
+
+JMJD6 forms an interaction network with proteins involved in splicing and protein synthesis
+
To understand the mechanism of JMJD6 in regulating splicing in neuroblastoma, we performed an unbiased identification of JMJD6 interacting partners by introducing a flag-tagged JMJD6 into SK-N-AS and BE2C cells, followed by immunoprecipitation to pull down the JMJD6 associated complex and protein identification with mass spectrometry (Figure 5a, Supplementary Table 7). We found that JMJD6 mainly interacted with two classes of proteins which are involved in RNA splicing and protein synthesis in both cell lines, respectively (Figure 5a, Supplementary Figure 6). We then validated the interactions of JMJD6 with splicing factors using immunoprecipitation and western blot, and demonstrated JMJD6 formed a complex with these RNA binding proteins, including RBM39 (Figure 5b), a therapeutic target of high-risk neuroblastoma27. Since JMJD6 also interacted with several molecules involved in protein translation, we investigated if JMJD6 also regulates protein synthesis by using an approach, Click-IT AHA to label the newly synthesized proteins, followed by western blot assessment (Figure 5c). Interestingly, overexpression of JMJD6 greatly reduced total protein synthesis (Figure 5c), suggesting that JMJD6 may antagonize protein production.
+
+
+
JMJD6 forms an interaction network consists of proteins involved in splicing and protein synthesis
+
a. Flag tagged JMJD6 transduced into SK-N-AS cells for immunoprecipitation with anti-Flag followed by protein identification by mass spectrometry. The interacting protein partners of JMJD6 are analyzed by STRING protein network.
+
b. Immunoprecipitation followed by western blot to validate the JMJD6 interacting partners in SK-N-AS cells. IP= immunoprecipitation, FT= flowthrough.
+
c. Click-IT AHA labeling showing the newly synthesized proteins after overexpression of JMJD6 in SK-N-AS cells.
+
d,e. Western blot showing the expression of GAC and KGA isoforms in BE2C (d) SK-N-AS (e), after U2AF2 and CPSF6 knockdown for 72hours.
+
+
+
+
Then we determined if the splicing factors interacting with JMJD6 also regulate GLS isoform expression. Among the splicing factors with which JMJD6 interacted, CPSF6 has been previously shown to regulate the alternative splicing of GLS78. We validated the function of CPSF6 in neuroblastoma cells and found that, indeed, loss of CPSF6 led to a dramatic switch from GAC to KGA in BE2C and SK-N-AS cells (Figure 5d, 5e). Previous studies showed that JMJD6 and U2AF2 (U2AF65) interact to regulate splicing50,79. We also found that knockdown of U2AF2 resulted in a similar phenotype to JMJD6 knockdown in that the expression of KGA isoform was greatly increased in both cell lines (Figure 5d, 5e). These data collectively support the functions of JMJD6 in regulating the splicing of metabolic genes and protein homeostasis in MYC-driven neuroblastoma.
+
+
+JMJD6 regulates production of TCA intermediates and nucleoside triphosphate
+
To further dissect the biological consequences of loss of function of JMJD6, we created stable JMJD6 knockout clones (Supplementary Figure 7) and defined the metabolite spectrum affected by loss of function by using liquid chromatography with tandem mass spectrometry (LC-MS/MS). JMJD6 knockout greatly reduced the production of tricarboxylic acid cycle (TCA) intermediates (i.e., oxoglutarate, fumarate) and nucleoside triphosphate (ATP, CTP, GTP) (Figure 6a), indicating that JMJD6 is a key bioenergetics regulator in cancer cells. Pathway analysis revealed that the reduced metabolites were involved in the Warburg effect, TCA, pentose phosphate pathway, and mitochondrial electron transport chain (Figure 6b), all of which are critical for providing cancer cell bioenergetics for proliferation and survival. Oxoglutarate (α-ketoglutarate) and fumarate are downstream products of glutaminolysis (Figure 6c). We reasoned that cellular metabolites such as glutamate and oxoglutarate may predict the cytotoxic effect of loss-of-function of JMJD6. If cells have higher levels of glutamate and oxoglutarate, they might be less dependent on JMJD6 due to their higher capacity of buffering against reduced glutaminolysis. To test this hypothesis, we used DepMap data that included 225 metabolites in 928 cancer cell lines from over 20 lineages80, and analyzed the correlation of each metabolite with JMJD6 gene dependency. The data showed that cells with high levels of AMP, glutamate, alanine, 2-hydroxyglutarate and 2-oxoglutarate were more resistant to JMJD6 knockout, while cells with high levels of lactose and sucrose were more sensitive to JMJD6 knockout (Figure 6d, Supplementary Table 8). High levels of AMP activate AMP kinase (AMPK), consequently leading to enhanced fatty acid oxidation to stimulate ATP production while alanine can be converted to pyruvate to provide acetyl-CoA to fuel the TCA cycle (Figure 6c). Therefore, high levels of AMP and alanine may provide cells alternative bioenergetics sources for survival. These data further indicate that JMDJ6 function is wired into the regulation of mitochondrial metabolism.
+
+
+
JMJD6 regulates production of citric acid cycle intermediates and NTP
+
a. Heatmap showing the metabolites differentially expressed in SK-N-AS cells (n=5) after JMJD6 knockout (n=5) based on LC-MS/MS analysis.
+
b. Pathway analysis of metabolites downregulated by JMJD6 knockout.
+
c. Pathway cartoon showing the connections of TCA, glycolysis, glutaminolysis, and β-oxidation.
+
d. Correlation of metabolite abundance with JMJD6 dependency. The positive correlation indicates that the higher the abundance of metabolites, the more resistance of cells to JMJD6 knockout. On the contrary, the negative correlation indicates the higher the abundance of metabolites, the more sensitive of cells to JMJD6 knockout.
+
+
+
+
+
+JMJD6 determines the efficacy of indisulam, a molecular glue degrading splicing factor RBM39
+
Dysregulated splicing as a vulnerability of MYC-driven cancers provides a rationale to target neuroblastoma by using splicing inhibitors as a therapeutic approach. We and others have recently reported that indisulam, a “molecular glue” that selectively degrades the splicing factor RBM39, is exceptionally effective at causing tumor regression in multiple high-risk neuroblastoma models without overt toxicity27,28, suggesting indisulam has translational potential. Understanding the factors determining the efficacy of indisulam or any other drug is critical for developing precision therapy, combination therapy or preventing therapy resistance. In addition to complexing together (Figure 5a, b), JMJD6 and RBM39 exhibit significant correlation of co-dependency in cancer cells, namely, cancer cells have similar dependency on JMJD6 and RBM39 for survival (Supplementary Table 3, Figure 7a). These data indicate that JMJD6 may play a role in modulating the effect of indisulam. To test this hypothesis, we performed GSEA to identify the differential pathways between indisulam-sensitive and indisulam-less sensitive neuroblastoma cells. It turned out that histone lysine demethylase (HDM) genes including JMJD6 are the in the only gene signature that is significantly enriched in indisulam-sensitive cells (Figure 7b, c, d). Indeed, knockout of JMJD6 led to partial but significant resistance to indisulam treatment of SK-N-AS cells (Figure 7e-g) and BE2C cells (Figure 7h-j), supporting that cells with high JMJD6 expression are more dependent on RBM39. Since we found that JMJD6 plays a key role in modulating glutaminolysis, we tested if expression of GAC or KGA could affect the activity of indisulam. Interestingly, overexpression of either GAC and KGA renders BE2C and SK-N-AS cells more resistant to indisulam treatment (Figure 7k-n), suggesting that enhanced glutaminolysis may confer therapeutic resistance to spliceosome inhibition.
+
+
+
JMJD6-GAC pathway regulates the response of neuroblastoma cells to indisulam treatment
+
a. Spearman correlation of effects of JMJD6 knockout and RBM39 knockout demonstrating the co-dependency of JMJD6 and RBM39 from DepMAP CRISPR screening data (n=1086). Each dot represents one cell line.
+
b. GSEA analysis for indisulam sensitive vs resistant neuroblastoma cell lines based on CTD2 (Cancer Target Discovery and Development) data showing histone lysine demethylase gene signature is the one that is significantly associated with indisulam response.
+
c. Heatmap from GSEA (b) showing the individual genes in indisulam sensitive and resistant cells.
+
d. JMJD6 expression in indisulam sensitive and resistant neuroblastoma cells. p value calculated by student t test.
+
e. Western blot showing JMJD6 knockout in SK-N-AS cells using indicated antibodies.
+
f. Colony formation of SK-N-AS cells in triplicates with or without JMJD6 knockout treated with different concentrations of indisulam for 7 days, stained with crystal violet.
+
g. Quantification of cell density by using Image J from f. ns=not significant. ** p<0.001, ***p<0.0001
+
h. Western blot showing JMJD6 knockout in BE2C cells using indicated antibodies.
+
i. Colony formation of BE2C cells in triplicates with or without JMJD6 knockout treated with 100nM of indisulam for 5 days, stained with crystal violet.
+
j. Quantification of cell density by using Image J from f. ns=not significant. ** p<0.001, ***p<0.0001
+
k. Colony formation of BE2C cells in triplicates with KGA and GAC overexpression treated with 250nM of indisulam for 5 days, stained with crystal violet.
+
l. Colony formation of SK-N-AS cells in triplicates with KGA and GAC overexpression treated with 100nM of indisulam for 7 days, stained with crystal violet.
+
m. Quantification of cell density by using Image k from k. * p<0.01, **p<0.001
+
n. Quantification of cell density by using Image l from k. **p<0.001
+
+
+
+
+
+
+Discussion
+
MYC is an oncogenic driver of many types of cancer and plays a pivotal role in regulating glycolysis, glutaminolysis, nucleotide and lipid synthesis, and ribosome and mitochondrial biogenesis6. Recent studies have revealed that there is also an interplay between MYC and pre-mRNA splicing machinery7-10,81-84. Pre-mRNA splicing is an essential biological process catalyzed by the spliceosome to produce mature mRNAs85,86. Over 90% of multiexon genes in the human genome undergo alternative splicing87,88, which significantly expands the diversity of the proteome and consequently impacts various biological functions89,90. In this study, we found that the chromosome 17q locus, which is frequently gained in MYC-driven cancers, houses numerous genes essential to cancer cell survival that are implicated in pre-mRNA splicing, ribosome and mitochondrial biogenesis and other biological functions. Particularly, JMJD6, which is located on 17q25 and is amplified in a number of cancer types, physically interacts with a subset of splicing factors such as RBM39 and regulates the alternative splicing of metabolic genes. Depletion of JMJD6 inhibits cancer cell proliferation and impedes tumor growth while overexpression of JMJD6 promotes MYC-mediated tumorigenesis, suggesting that JMJD6 and potentially other 17q genes have oncogenic functions in cellular transformation. Previous studies suggest that JMJD6 and BRD4 interact to regulate gene transcription56,66. However, our unbiased identification of the JMJD6 interactome only identified a subset of proteins involved in mRNA splicing and protein translation in neuroblastoma cells, suggesting that JMD6 may predominantly regulate protein homeostasis to facilitate MYC-mediated transformation. Nevertheless, we cannot exclude the possibility that our immunoprecipitation conditions were too harsh, leading to dissociation of proteins that loosely or dynamically bind to JMJD6.
+
MYC-induced metabolic reprogramming triggers cellular dependency on exogenous glutamine as a source of carbon for mitochondrial membrane potential maintenance and macromolecular synthesis2, leading to “glutamine addiction”2. Glutaminolysis is a process by which GLS and GLS2 convert glutamine to glutamate, which is, in turn, converted by glutamate dehydrogenase or transaminase to 2-oxoglutarate that is further catabolized in the TCA cycle. Additionally, glutamate is a substrate for production of glutathione, an important antioxidant. We previously showed that neuroblastoma relies on MYCN-induced glutaminolysis for survival20. In this study, our RNA-seq barely detected the expression of GLS2 in the neuroblastoma cell models we used, indicating that GLS is the major enzyme that catalyzes glutamine in these model systems. GLS has two isoforms, GAC and KGA, resulting from alternative splicing. KGA is mainly localized in the cytoplasm while GAC is localized in mitochondria and has a higher basal activity73. GAC mRNA levels strongly correlate with the conversion of glutamine to glutamate, as a proxy for GAC activity91. The positive correlation of JMJD6 and GAC suggests that the JMJD6-high tumors have enhanced glutaminolysis activity. CSPF6 and the noncoding RNA CCAT2 have been reported to regulate the splicing of GLS isoforms78,92. Interestingly, we found that depletion of JMJD6 leads to a GLS isoform switch from GAC to KGA, indicating that JMJD6 is involved in alternative splicing of GLS. We further found that JMJD6 physically interacts with CPSF6 in a splicing network and validated that loss of CPSF6 results in remarkable induction of the KGA isoform. We further validated that other splicing factors such as U2AF2 also interact with JMJD6 to regulate the GLS isoform switch. These data indicate that cancer cells can adjust metabolism through alternative splicing to produce enzymes with distinct subcellular localization and activity that promote cellular transformation or progression of an oncogenic phenotype. The cooperation of JMJD6 and MYC in cellular transformation further supports the hypothesis that JMJD6 is needed for metabolic reprogramming triggered by MYC. However, overexpression of either GAC or KGA promotes cell proliferation, suggesting that the switching of KGA/GAC is a cellular fitness mechanism in response to interruption of the spliceosome by adjusting the metabolic rate. Within the tumor microenvironment (i.e., replete and deplete oxygen and nutrient supply) GLS activity is possibly finely tuned through splicing mechanism for adaption.
+
Additionally, we found that JMJD6 physically interacts with a subset of ribosomal proteins that are responsible for protein translation. Interestingly, overexpression of JMJD6 reduces global protein synthesis. A recent study showed that MYC overactivation leads to proteotoxic stress in cells by enhancing global protein synthesis, consequently causing cell death93. The increased global protein synthesis by MYC needs to be buffered through loss of DDX3X, a regulator of ribosome biogenesis and global protein synthesis, for lymphomagenesis93. Our findings suggest that, besides the functions in regulating alternative splicing for metabolism, JMJD6 is involved in MYC-mediated cell transformation by buffering unwanted proteotoxic stress due to high rate of protein synthesis induced by MYC (Figure 8).
+
+
+
Working mechanism of JMJD6 in MYC-driven neuroblastoma.
+
Overactive MYC drives high-load of gene transcription, enhanced protein synthesis and high rate of metabolism, leading to detrimental cellular stresses and consequent cell death (Model a). However, when 17q is amplified, high levels of JMJD6 and other proteins encoded by 17q genes physically interacts with the splicing and translational machineries, enhancing pre-mRNA splicing of metabolic genes such as GLS and inhibiting global protein synthesis, respectively, leading to reduced detrimental stresses and enhanced cancer cell survival and tumorigenesis (Model b). The high levels of JMJD6 predicts high dependency of RBM39, which are more sensitive to indisulam treatment.
+
+
+
+
Neuroblastoma is responsible for as much as 15% of childhood cancer mortality94. With current intensive multimodal therapies, 5-year survival rates for high-risk patients remain less than 50%95-98. In addition, survivors of high-risk disease have a significant risk of developing long-term side effects including subsequent malignant neoplasms due to cytotoxic chemotherapy and radiotherapy99,100. Unfortunately, developing effective precision therapies against high-risk neuroblastoma has been challenging due to the lack of targetable recurrent mutations in neuroblastoma29,30,101. Our previous study showed that indisulam, the splicing inhibitor that targets RBM39, a JMJD6-interacting partner, induced a durable complete response in multiple high-risk neuroblastoma models, supporting its potential use in future clinical trials. Our current study showed that JMJD6 expression is positively correlated with the effect of indisulam, and knockout of JMJD6 confers resistance to indisulam treatment. In line with the biological functions of JMDJ6 in regulating GLS isoform expression and mitochondrial metabolism, overexpression of GAC or KGA also caused resistance to indisulam treatment. These data indicate that JMJD6 could serve as a biomarker that predicts response to indisulam or other splicing inhibitors.
+
+
+Materials and Methods
+
+Cell lines
+
KELLY, SIMA, BE2C, IMR32, SK-N-AS, CHLA20 were cultured in 1X RPMI1640 (Corning, 15-040-CV) supplemented with 10% Fetal Bovine Serum (Sigma-Aldrich, F2442), 1% L-Glutamine (Corning, A2916801). NIH3T3 and 293T cells were cultured in 1X DMEM supplemented with 10% Fetal Bovine Serum (Sigma-Aldrich, F2442), 1% L-Glutamine (Corning, A2916801). All cells were maintained at 37 °C in an atmosphere of 5% CO2. JoMa1 cells were kindly provided by Dr. Schulte (Department of Pediatric Oncology and Hematology, University Children’s Hospital Essen, Essen, Germany) were cultured in NCC Medium: DMEM (4.5 mg/ml Glucose, L-Glutamine, Pyruvate): Ham’s F12 (1:1) was supplemented with: 1% N2-Supplement (Invitrogen, no. 17502-048), 2% B27-Supplement (Invitrogen, no. 17504-044), 10ng/ml EGF (Invitrogen), 1ng/ml FGF (Invitrogen), 100U/ml Penicillin–Streptomycin (Invitrogen) and 10% Chick-Embryo-Extract (Gemini Bio-Products, CA). Neural crest culture medium was supplemented with 200 nM 4-OH-tamoxifen (Sigma no. H7904) in routine culture to ensure nuclear localization of c-MycERT and JoMa1 cell proliferation. JoMa1 cells were grown on cell culture flask/dish coated with fibronectin, NCC-medium supplemented with 200nm 4-OHT was changed daily. Cells were passaged after 3–4 days in culture when 70% confluence was reached (4 X 106 cells/10 cm dish).
+
All human-derived cell lines were validated by short tandem repeat (STR) profiling using PowerPlex® 16 HS System (Promega) once a month. Additionally, a polymerase chain reaction (PCR)-based method was used to screen for mycoplasma once a month employing the LookOut® Mycoplasma PCR Detection Kit (MP0035, Sigma-Aldrich) and JumpStart™ Taq DNA Polymerase (D9307, Sigma-Aldrich) to ensure cells were free of mycoplasma contamination.
+
+
+Retroviral plasmids and retrovirus packaging
+
MSCV-IRES-GFP and MSCV-IRES-mCherry were obtained from St Jude Vector Core. Human JMJD6 and murine MYCN were subcloned into MSCV-IRES-GFP and MSCV-IRES-mCherry, respectively. The MSCV-CMV-CMV-Flag-HA-JMJD6 was purchased from Addgene (Addgene # plasmid 31358). The retrovirus packaging was done as described in the following procedure. Briefly, HEK93T cells were transfected with viral vectors by combining 5μg of target vector, 4.4μg of pMD-old-gag-pol, and 0.6μg of VSV-G plasmids in 400uL of DMEM without serum or L-glutamine. PEIpro transfection reagent (Polyplus 115-010) was added at 2:1 (PEIpro μL: μg of plasmid) per 100mm dish of cells and mixed well, and incubated at RT for at least 20 minutes, prior to adding cells. The following day, fresh medium was added to cells. For 3-4 days, viral media was harvested and replaced twice per day. Viral media was centrifuged at 1500RPM for 10 minutes and filtered through a 0.45um vacuum filter. Virus was concentrated by ultracentrifugation at 28.5kRPM for 2 hours at 4C, aspirated, and resuspended in either OptiMEM or PBS, aliquoted, and frozen at -80C until use. Wasie, add information for GAC/KGA here.
+
+
+siRNA transfection
+
25uM of each siRNA oligo was resuspended in 500uL of prewarmed Opti-MEM, reduced serum medium (Gibco Life technologies #31985-070) in 6 well plates. To each well, 7μL of RNAiMax (Invitrogen Lipofectamine RNAiMAX transfection reagent 13778100) was added, mixed, and left at room temperature for 10 minutes. After incubation, 100,000 cells of each indicated cell line were added to each well in a total of 2mL volume with RPMI medium supplemented with 10% FBS. JMJD6 siRNA#43, 5-CCAAAGUUAUCAAGGAAA-3; JMJD6 siRNA#45, 5-CAGUGAAGAUGAAGAUGAA-3. U2AF2 siRNA#1 AGAAGAAGAAGGUCCGU; U2AF2 siRNA#2 GUGGCAGUUUCAUAUUUG. CPSF6 siRNA#1 GGAUCACCUUCCAAGACA. CPSF6 siRNA#2 AGAACCGUCAUGACGAUU.
+
+
+SDS-PAGE and Western blot
+
Cells were washed twice with ice-cold phosphate-buffered saline (PBS) and directly lysed on ice with 2X sample loading buffer (0.1 M Tris HCl [pH 6.8], 200 mM dithiothreitol [DTT], 0.01% bromophenol blue, 4% sodium dodecyl sulfate [SDS] and 20% glycerol). On ice, cell lysates were sonicated once with a 5 second bursts at 40% amplitude output (Sonics, VIBRA CELL) followed by 25 minutes heating at 95 °C. After the cell lysates were centrifuged at 13,000 × g at room temperature for 2 mins, 10-20 µl of the cell lysates were separated on 4-15% Mini-PROTEAN® TGX™ Stain-FreeTM Protein Gels from Bio-Rad and transferred to methanol-soaked polyvinylidene difluoride (PVDF) membranes (Millipore). Lysates for RBM39 G268V mutant cell lines and DCAF15 genetically modified cells were generated as previously described102. Membranes were blocked in PBS buffer supplemented with 0.1% TWEEN 20 and 5% skim milk (PBS-T) and incubated for 1 hour at room temperature under gentle horizontal shaking. Membranes were incubated overnight at 4 °C with the primary antibodies. The next day, membranes were washed 3 times (for 5 minutes) with PBS-T at room temperature. Protected from light, membranes were then incubated with goat anti-mouse or goat anti-rabbit HRP-conjugated secondary antibodies (1:5,000) for 1 hour at room temperature, followed by three 5-minite washes with PBS-T at room temperature. Lastly, membranes were incubated for 1 minute at room temperature with SuperSignal West Pico PLUS Chemiluminescent Substrate (34580, Thermo Fisher Scientific) and the bound antigen-antibody complexes were visualized using Odyssey Fc Imaging System (LI-COR Corp., Lincoln, NE).
+
+
+RNA extraction and RT-PCR isoforms of GLS
+
RNA was extracted using RNeasy® Plus Mini Kit (Qiagen, reference # 74136) following the manufacturer’s protocol. cDNA was prepared in 20ul reaction from 500ng of total RNA using Superscript™ IV First Strand Synthesis System (Invitrogen, reference # 1809105) kit. Real-time PCR reactions were run in triplicates (n=3) in the 7500 Real-time PCR system by Applied Biosystems (Thermo Fisher Scientific) using power SYBR Green PCR master mix (Applied Biosystems, reference # 4367660). ΔΔCT methods were applied to analyze the results. The following primers were used to perform the quantitative Real-time PCR-GAPDH (Forward: AACGGGAAGCTTGTCATCAATGGAAA, Reverse: GCATCAGCAGAGGGGGCAGAG), GAC (Forward: GAGGTGCTGGCCAAAAAGCCT, Reverse: AGGCATTCGGTTGCCCAAACT), KGA (Forward: CTGCAGAGGGTCATGTTGAA, Reverse: ATCCATGGGAGTGTTATTCCA).
+
+
+Lentiviral packaging of pLenti and shRNA
+
The GAC and KGA cDNAs were synthesized by Genscript company and cloned into pLenti vector. The TRC lentiviral-based shRNA knockdown plasmids for JMJD6 were purchased from Horizon Discovery (sh#46: RHS3979-201781036, TTAAACCAGGTAATAGCTTCG; sh#47: RHS3979-201781037, ATCTTCACTGAGTAGCCATCG) The lentiviral shJMJD6 and shControl (pLKO.1) particles were packaged by Vector Lab at St Jude. Briefly, HEK293T cells were transfected with shRNA constructs and helper plasmids (pCAG-kGP1-1R, pCAG4-RTR2, and pHDM-G). The 48- and 72-hr post-transfection replication-incompetent lentiviral particles were harvested and transduced into cells with 8μg/ml of polybrene. 48 hours later, 1 μg/ml of puromycin was added for selection for additional 48 hours before injection into mice or immunoblotting.
+
+
+JMJD6 CRISPR KO method
+
Genetically modified neuroblastoma cells were generated by using CRISPR-Cas9 technology. Briefly, 400,000 NB cells were transiently co-transfected with 100pmol of chemically modified gRNA (GGACTCTGGAGCGCCTAAAA) (Synthego), 33pmol of Cas9 protein (St. Jude Protein Production Core), 200ng of pMaxGFP (Lonza), and, using solution P3 and program DS-150 in small cuvettes according to the manufacturer’s recommended protocol. Five days post-nucleofection, cells were sorted for GFP+ (transfected) cells and plated as single cells into 96-well plates. Cells were clonally expanded and screened for the desired modification using targeted next generation sequencing followed by analysis with CRIS.py (https://pubmed.ncbi.nlm.nih.gov/30862905/).
+
+
+Colony formation for JoMa1 cells
+
Matrigel was kept at 4°C to being liquified for 6 hours. 50μL of Matrigel per 1 square centimeter area was added to 24-well plate without air bubble. The 24-well plate was kept at 37°C in cell culture incubator till it was solidified. 200 of JoMa1 cells transduced with GFP, JMJD6, MYCN, MYCN+JMJD6 in DMEM:F12 enriched media without tamoxifen were seeded onto the 24-well coated with Matrigel. This was done in triplicate. Cells were checked daily, and media were changed every 3 days without disturbing the Matrigel by removing and adding media gently. To stain the colonies, cells were fixed by formaldehyde (3.7% in PBS) for 2 min at room temperature, followed by permeabilization with 100% methanol (not ice-cold) for 20min at room temperature. The colonies were stained by 0.4% crystal violet.
+
+
+Crystal Violet Staining
+
After removing media, cells were washed with Dulbecco’s phosphate buffered saline without calcium or magnesium (DPBS, Lonza) and treated with 4% formaldehyde in PBS (PFA) for 20 minutes. Once PFA was removed, cells were stained with 0.1% crystal violet stain for 1 hour. KGA/GAC overexpression colony formation: 5000 cells were plated of BE2C control, KGA and GAC overexpressing cells and were culture for 7 days; 10,000 cells were plated of SKNAS control, KGA and GAC overexpressing cells and were culture for 7 days (n=3). After 7 days, medium was removed and cells were washed with Dulbecco’s PBS (DPBS) (DPBS, Lonza) and treated with 4% formaldehyde in PBS [paraformaldehyde (PFA)] for 30 min. PFA was later removed and cells were stained with 0.1% crystal violet stain for 1 hour. Experiments were repeated twice. Indisulam treatment on WT and JMJD6-KO cell lines: JMJD6-WT and KO cells were plated in 12-well plate (50,000 cells/well for SKNAS) and 6-well plate (5,000 cells/well for BE2C) (n=3). Next day, cells were treated with Indisulam with indicated concentration for 7 days (SKNAS cells) and 5 days (BE2C cells). Crystal violet staining was performed to visualize and quantify the colony formation. Experiments were repeated twice. Indisulam treatment on KGA/GAC overexpressing cell lines: 10,000 BE2C and 100,000 SKNAS control, KGA and GAC overexpressing cells were plated in 6-well plate (n=3). Next day, cells were treated with Indisulam for 7 days. Crystal violet staining was performed to visualize and quantify the colony formation. Experiments were repeated twice.
+
+
+Click-iT AHA labeling assay for metabolic labeling of newly synthesized proteins
+
Click-iT was performed as previously described. Briefly, cells were plated at 5 million cells per 100mm dish in RPMI supplemented with 10% FBS. Cells were washed with warm PBS and replaced with methionine-free medium (Thermo 21013024 supplement with glutamine and sodium pyruvate) for 1 hour at 37° C in 5% CO2. Following, fresh methionine-free media containing 50μM of Click-iT AHA (L-azidohomoalanine) (Thermo C10102) was added to the cells for 2 hours at 37°C. After AHA-labeling, cells were washed with warm PBS and lysed with 1% SDS, 50mM Tris-HCl, (pH 8.0) supplemented with phosphatase inhibitors (PhosSTOP, Sigma) and protease inhibitors (cOmplete mini, Roche) by applying the buffer directly to the plate, incubating the cells on ice for 30 minutes, tilting the plates, and collecting the lysate. Lysates were briefly sonicated, vortexed for 5 minutes, and centrifuged at 18,000xg for 5 minutes at 4°C. Total protein quantification was assayed using the EZQ Protein Quantification Kit (Thermo R33200) according to the manufacturer’s protocol and results were read on a fluorescence-based microplate reader (BioTek Synergy 2). Click chemistry of the biotin-alkyne (PEG4 carboxamide-propargyl biotin) (Thermo B10185) to the AHA-labeled lysates was performed using the Click-iT Protein Reaction Buffer Kit (Thermo C10276) using a concentration of 40μM biotin-alkyne per click reaction (and no biotin-alkyne added for controls). Following the click reaction, samples were either assayed for total biotinylated protein by following the manufacturer’s protocol. For total biotinylated protein, briefly, 600μL of methanol, 150μL of chloroform, and 400μL of megaOhm water was sequentially added and vortexed, followed by centrifugation at 18,000xg for 5 minutes. The upper aqueous phase was discarded, and 450μL of methanol was added, vortexed, and centrifuged again at 18,000xg for 5 minutes. This methanol step was performed in duplicate to remove residual reaction components. Protein pellets were allowed to air dry and resuspended in a suitable volume of sample buffer and heated prior to western blot analysis.
+
+
+Immunoprecipitation
+
5X106 BE2C and SK-N-AS cells expressing Flag-JMJD6 were cultured in 150cm dish with complete RPMI media. Cells were washed twice with cold PBS after reaching 95% confluency, then lysed in 1ml lysis buffer (50mM Tris-HCl, pH 7.4, 150mM NaCl, 1mM EDTA, 1% Trion-X100 with complete protease inhibitors (Sigma 11836170001, added fresh) and PhosSTOP (Sigma 4906845001). Cells were scrapped into a 1.5mL Eppendorf tube and incubated on ice for 15min, which were mixed by vortex every 5 min. Cell lysates were spun by 135000rpm for 10min at 4C. The supernatant was transferred to a new tube. The cell lysates were subject to immunoprecipitation using M2 anti-Flag beads (Sigma, M8823) overnight by rocking at 4°C. The following day, beads were washed 3X with buffer and eluted with 5 packed gel volumes of FLAG peptide in TBS buffer (3uL of stock FLAG peptide at 5ug/uL per 100uL of TBS buffer) while rotating at 4°C for 30 minutes. Beads were briefly spun and the supernatant was removed from the beads (eluate). This elution step was repeated one more time and pooled with the first eluate. Prior to western blot, input, flow throughs, and elution samples were processed by adding 4X sample buffer supplemented with 50mM DTT and heated at 75°C for 10 minutes prior to running on a gel.
+
+
+RNA-immunoprecipitation
+
SK-N-AS cells expressing Flag-JMJD6 were grown in a 10-cm dish in RPMI complete media. After 70% confluency, cells were washed with cold PBS twice and then were subject to lysis with Polysome Lysis Buffer (100mM KCl, 5mM MgCl2, 10mM HEPES, pH7.0, 0.5% NP-40, 1mM DTT, 100 U/ml RAasin RNase inhibitor (Promega, N2511), 2mM vanadyl ribonucleoside complexes solution (Sigma, 94742), 25μL protease inhibitor cocktail for mammalian cells (Sigma, P8340)). Cell lysates were precleared with magnetic IgG beads for 1 hour. The cell lysates were subject to immunoprecipitation using M2 anti-Flag beads (Sigma, M8823) overnight by rocking at 4°C. The same amounts of lysates were saved at -80°C for input RNA extraction. The beads were washed with 250 μL Polysome Lysis Buffer for 4 times, flowed by washing with Polysome Lysis Buffer containing 1M urea. RNA was released by adding 150μL of Polysome Lysis Buffer containing 0.1% SDS and 45μg protease K (Ambion, AM2548) and incubated at 50°C for 30min. RNA was extracted with phenol-chloroform-isoamyl alcohol mixture (Sigma, 77618). RNA was recovered by adding 2μL of GlycoBlue (15mg/ml, Ambion, AM9516), 36μL of 3M sodium acetate and 750μL ethanol followed by incubation at -20°C for overnight. RNA was precipitated with 70% ethanol and air dried, followed by resuspension with RNase-free water followed by DNaseI (Promega, M6101) treatment to remove genomic DNA. The resultant RNAs were subjected to RT-qPCR analysis using 3 sets of GAC and KGA primers and 18S rRNA as control. 18S rRNA F: GCTTAATTTGACTCAACACGGGA; 18S rRNA R: AGCTATCAATCTGTCAATCCTGTC. GLS-GACiso_F: GAGGTGCTGGCCAAAAAGCCT; GLS-GACiso_R: AGGCATTCGGTTGCCCAAACT. GLS-KGAiso_F: CTGCAGAGGGTCATGTTGAA; GLS-KGAiso_R: ATCCATGGGAGTGTTATTCCA. KGA_set2_F: GCAGCCTCCAGGTGCTTTCA; KGA_set2_R: GTAATGGGAGGGCAGTGGCA. KGA_set3_F: TGCCCGACACTGCCCTTTAG; KGA_set3_R: CCTGCCAGACAGACAACAGCA. GAC_set2_F: TGCTTCTCAAGGCCTTACTGC; GAC_set2_R: AGGCATTCGGTTGCCCAAACT. GAC_set3_F: CCTTCTAGAGGTGCTGGCCAAA; GAC_set3_R: TGCAACACAAATATGCAGTAAGGC. For validation of protein immunoprecipitation, 20% of beads after overnight incubation were removed and processed as follows: Beads were washed 3X with buffer and eluted with 5 packed gel volumes of FLAG peptide in TBS buffer (3μL of stock FLAG peptide at 5μg/μL per 100uL of TBS buffer) while rotating at 4°C for 30 minutes. Beads were briefly spun and the supernatant was removed from the beads (eluate). This elution step was repeated one more time and pooled with the first eluate.
+
+
+Identification of JMJD6 interacting partners by LC-MS/MS
+
Protein samples were run on a short gel as described in a previously published protocol103. Proteins in the gel bands were reduced with dithiothreitol (DTT) (Sigma) and alkylated by iodoacetamide (IAA) (Sigma). The gel bands were then washed, dried, and rehydrated with a buffer containing trypsin (Promega). Samples were digested overnight, acidified and the resulting peptides were extracted. The extracts were dried and reconstituted in 5% formic acid. The peptide samples were loaded on a nanoscale capillary reverse phase C18 column by a HPLC system (Thermo EASY-nLC 1000) and eluted by a gradient. The eluted peptides were ionized and detected by a mass spectrometer (Thermo LTQ Orbitrap Elite). The MS and MS/MS spectra were collected over a 90-min liquid chromatography gradient. Database searches were performed using Sequest (version 28 revision 13) search engine against a composite target / decoy Uniprot human protein database. All matched MS/MS spectra were filtered by mass accuracy and matching scores to reduce protein false discovery rate to <1%. Spectral counts, matching to individual proteins reflect their relative abundance in one sample after the protein size is normalized. The spectral counts between samples for a given protein was used to calculate the p-value based on G-test 104.
+
+
+Metabolome profiling by LC-MS/MS
+
MJD6 knockout or parental SK-N-AS cells were cultured in 6-well plates to ∼85% confluence and washed with 2 mL ice cold 1X Phosphate-Buffered Saline (PBS). The cells were then harvested in 300 µL freezing 80% acetonitrile (v/v) into 1.5 mL tubes and lysed in the presence of 0.5mm Zirconia/silica beads by Bullet Blender (Next Advance) at 4 °C until the sample were homogenized. The resulting lysate was then centrifuged at 21,000 x g for 5 min and the supernatant was dried by speedvac. The samples were resuspended in 50 µL of 1% acetonitrile plus 0.1% trifluoroacetic acid, and separated by Ultra-C18 Micro spin columns (Harvard apparatus) into hydrophilic metabolites (flow through) and hydrophobic metabolites (eluent of 125 µL of 80% acetonitrile plus 0.1% trifluoroacetic acid). Ten µL of hydrophilic metabolites were dried, reconstituted in 3 µL of 66% acetonitrile and analyzed by a ZIC-HILIC column (150 × 2.1 mm, EMD Millipore) coupled with a Q Exactive HF Orbitrap MS (Thermo Fisher) in negative mode and metabolites were eluted within a 45 min gradient (buffer A: 10mM ammonium acetate in 90% acetonitrile (pH=8); buffer B: 10mM ammonium acetate in 100% H2O (pH=8)). Twenty µL of hydrophobic metabolites were dried and resuspend in 3 µL of 5% formic acid followed by separation with a self-packed nanoC18 column (75 μm × 15 cm with 1.9 µm C18 resin from Dr. Maisch GmbH) and detection with a Q Exactive HF Orbitrap MS (Thermo Fisher) in positive mode. Metabolites were eluted within a 50 min gradient (buffer A: 0.2% formic acid in H2O; buffer B: 0.2% formic acid in acetonitrile). MS settings for both types of samples included MS1 scans (120,000 resolution, 100-1000 m/z, 3 x 106 AGC and 50 ms maximal ion time) and 20 data-dependent MS2 scans (30,000 resolution, 2 x 105 AGC, ∼45 ms maximal ion time, HCD, Stepped NCE (50, 100, 150), and 20 s dynamic exclusion). A mix of all samples served as quality control was injected in the beginning, middle and the end of the samples to monitor the signal stability of the instrument. The data analysis was performed by a recently developed software suite JUMPm. Raw files were converted to mzXML format followed by peak feature detection for individual sample and feature alignment across samples. Metabolite identification was supported by matching the retention time, accurate mass/charge (m/z) ratio, and MS/MS fragmentation data to our in-house authentic compound library and the matching of m/z and MS/MS fragmentation data to, downloaded experimental MS/MS library (MoNA, https://mona.fiehnlab.ucdavis.edu/), in-silico database generated from Human Metabolome Database (HMDB), and mzCloud (https://mzcloud.org). Peak intensities were used for metabolite quantification. The data was normalized by both cell numbers (before data collection) and trimmed median intensity of all features across samples (post data collection).
+
+
+Differential gene expression and gene set enrichment analysis (GSEA) for RNA-seq experiments
+
Total RNA from cells and tumor tissues were performed using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Paired-end sequencing was performed using the High-Seq platform with 100bp read length. Total stranded RNA sequencing data were processed by the internal AutoMapper pipeline. Briefly the raw reads were firs trimmed (Trim-Galore version 0.60), mapped to human genome assembly (GRCh38) (STAR v2.7) and then the gene level values were quantified (RSEM v1.31) based on GENCODE annotation (v31). Low count genes were removed from analysis using a CPM cutoff corresponding to a count of 10 reads and only confidently annotated (level 1 and 2 gene annotation) and protein-coding genes are used for differential expression analysis. Normalization factors were generated using the TMM method, counts were then transformed using voom and transformed counts were analyzed using the lmFit and eBayes functions (R limma package version 3.42.2). The significantly up- and down-regulated genes were defined by at least 2-fold changes and adjusted p-value < 0.05. Then gene set enrichment analysis (GSEA) was carried out using gene-level log2 fold changes from differential expression results against gene sets in the Molecular Signatures Database (MSigDB 6.2) (gsea2 version 2.2.3).
+
+
+RNA splicing analysis
+
After mapping RNA-seq data, rMATS v4.1.0 was used for RNA alternative splicing analysis by using the mapped BAM files as input. Specifically, five different kinds of alternative splicing events were identified, i.e., skipped exon (SE), alternative 5’-splicing site (A5SS), alternative 3’-splicing site (A3SS), mutually exclusive exon (MXE) and intron retention (RI). To keep consistent, the same GTF annotation reference file for mapping was used for rMATS. For stranded RNA-seq data, the argument “--libType fr-firststrand” was applied. To process reads with variable lengths, the argument “--variable-read-length” was also used for rMATS. To select statistically significantly differential splicing events, the following thresholds were used: FDR <0.05 and the absolute value of IncLevelDifference > 0.1. For visualization, the IGV Genome Browser was used to show the sashimi plots of splicing events. To investigate the genome-wide correlations of differential splicing between two genotypes (e.g., shRNA knockdown of JMJD6 and non-target shRNA in cells), we extracted splice junctions for all samples of both genotypes of interest from the STAR105 output files suffixed with “SJ.out.tab”, which contain high confidence collapsed splice junctions. Only those unique mapped reads crossing the junctions were considered. By extracting the union of the unique junction positions, we constructed a unified junction-read feature vector for each sample. Then, we normalized the junction-read vectors of each sample with TMM method in “voom” and “limma” and R package, assuming a negative binomial distribution. Next, we averaged the junction-read vectors for samples of the same genotype. The gene level expression was estimated based on the canonical junctions from the most abundant isoforms estimated for each gene. The fold changes of exon junctions significantly deviated from gene level changes were regarded as differentially spliced junctions for between cell-line comparisons.
+
+
+Data mining
+
JMJD6, GAC and KGA expression in tumor tissues were downloaded from R2 (https://portals.broadinstitute.org/ccle), Kocak dataset GSE45547 (649 samples) and Fischer dataset GSE120572 (394 samples). In both datasets, the probe UKv4_A_23_P311616 represented JMJD6, the probe UKv4_A_23_P308800 represented GAC and UKv4_A_23_P39766 represented KGA. JMJD6 expression data from the RNA-seq data of various pediatric cancer tissues were downloaded from St Jude cloud (https://pecan.stjude.cloud/). The copy number alterations of JMJD6 and the related Kaplan-Meier analysis were downloaded from cBioportal (cbioportal.org). The data for correlation of metabolite abundance and JMJD6 knockout effect were downloaded from DepMap (https://depmap.org/portal/).
+
+
+Pathway network analysis
+
The 114 essential fitness genes to neuroblastoma cell survival identified through genome-wide CRISPR/Cas9 library screen were uploaded into STRING program (https://string-db.org) for network interaction analysis with confidence threshold 0.15. The resulting network was then uploaded into Cytoscape program for presentation106. The clusters were grouped based on the biological functions of each gene.
+
Copy number analysis of JMJD6 and other genes encoding JmjC domain histone demethylases from St Jude neuroblastoma cohort.
+
Somatic copy number alternations (SCNA) were determined by CONSERTING (PMID: 25938371) for each pair of tumor and normal samples. The normalize read depth ratio (log2 ratio) for the CNV segments with JmjC-domain containing proteins were extracted and used for CNV heatmap generation (https://CRAN. R-project. org/package= pheatmap) and hierarchical clustering of samples.
+
+
+Xenograft studies
+
+
(1) shRNA-mediated JMJD6 knockdown. Neuroblastoma cells were transduced with shRNA lentiviral particles targeting JMJD6. 48 hours later, 1 μg/ml of puromycin was added for selection for additional 48 hours. Cancer cells (5x106) were mixed with Matrigel (1:1 ratio in volume) and subcutaneously injected into the flank sites of NSG mice. (2) JMJD6 and MYC-mediated transformation. After JoMa1 cells were transduced with GFP, JMJD6, MYCN and JMJD6/MYCN, 104 cells per group were mixed with Matrigel (1:1 ratio in volume) and subcutaneously injected into the flank sites of NSG mice. Mice were sacrificed when they reached the humane endpoint. Tumors were measured by using electronic calipers, and volumes calculated as width ρχ/6 xd3 where d is the mean of two diameters taken at right angles.
+
+
+
+Statistical analysis
+
All quantitative data are presented as mean ± SD. Unpaired Student’s t test was performed for comparison of two groups. Spearman correlation was used to assess the relationship between two variables. Kaplan-Meier method was used to estimate the survival rate. Mann-Whitney rank test (two-sided) was used to compare the tumor volume between two groups at every time point. P-values across multiple time points were adjusted for multiple comparison using the Holm-SidaK method. p<0.05 was considered as statistically significant. All the statistical analyses, except where otherwise noted, were performed using GraphPad Prism (v9).
+
+
+Data accessibility
+
GEO accession number: GSE185867 To review GEO accession GSE185867: Go to https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ncbi.nlm.nih.gov%2Fgeo%2Fquery%2Facc.cgi%3Facc%3DGSE185867&data=04%7C01%7Cjun.yang2%40stjude.org%7C56d8ac619531468173d608d98ea927d8%7C22340fa892264871b677d3b3e377af72%7C0%7C0%7C637697679446209542%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=XtZl1ttoOPBvfhwmX9uCcwVN1Yg02qDsvcsHdJv58xw%3D&reserved=0 Enter token qzixwsoohbsxjij into the box
+
+
+
+
+
+Acknowledgements
+
We thank the staff of the St. Jude Animal Resource Center and Hartwell Center for their dedication and expertise. The work was supported by American Cancer Society-Research Scholar (130421-RSG-17-071-01-TBG, J.Y.), National Cancer Institute (1R01CA229739-01, J.Y., R01CA266600, J.Y.). The work was also supported by the American Lebanese Syrian Associated Charities (ALSAC). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
+
+
+Author contributions
+
C.J., W.Q., S.S., H.H., D.H.B. D.H performed experiments. G.W., H.J., T-C.C., D.F., J-H.C performed proteomic, genomic and RNA-seq analysis. S.M.S and S.M.P-M provided CRISPR knockout. R.W., K.F. provided key reagents. A.M.D, A.M., J.P., and J.Y. supervised the studies. J.Y conceived the project and wrote the manuscript with help from co-authors.
+
+
+References
+Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell144, 646–674 (2011). 10.1016/j.cell.2011.02.013
+Pavlova, N. N. & Thompson, C. B. The Emerging Hallmarks of Cancer Metabolism. Cell Metab23, 27–47 (2016). 10.1016/j.cmet.2015.12.006
+Vazquez, A.et al.Cancer metabolism at a glance. J Cell Sci129, 3367–3373 (2016). 10.1242/jcs.181016
+DeBerardinis, R. J. & Chandel, N. S. Fundamentals of cancer metabolism. Sci Adv2, (2016). 10.1126/sciadv.1600200
+Kalkat, M.et al.MYC Deregulation in Primary Human Cancers. Genes (Basel) 8 (2017). 10.3390/genes8060151
+Stine, Z. E., Walton, Z. E., Altman, B. J., Hsieh, A. L. & Dang, C. V.MYC, Metabolism, and Cancer. Cancer Discov5, 1024–1039 (2015). 10.1158/2159-8290.CD-15-0507
+Hsu, T. Y.et al.The spliceosome is a therapeutic vulnerability in MYC-driven cancer. Nature525, 384–388 (2015). 10.1038/nature14985
+Hirsch, C. L.et al.Myc and SAGA rewire an alternative splicing network during early somatic cell reprogramming. Genes Dev29, 803–816 (2015). 10.1101/gad.255109.114
+Koh, C. M.et al.MYC regulates the core pre-mRNA splicing machinery as an essential step in lymphomagenesis. Nature523, 96–100 (2015). 10.1038/nature14351
+Phillips, J. W.et al.Pathway-guided analysis identifies Myc-dependent alternative pre-mRNA splicing in aggressive prostate cancers. Proc Natl Acad Sci U S A117, 5269–5279 (2020). 10.1073/pnas.1915975117
+Anczukow, O. & Krainer, A. R. The spliceosome, a potential Achilles heel of MYC-driven tumors. Genome Med7, (2015). 10.1186/s13073-015-0234-3
+Zhang, S.et al.MYCN controls an alternative RNA splicing program in high-risk metastatic neuroblastoma. Cancer Lett371, 214–224 (2016). 10.1016/j.canlet.2015.11.045
+Gustafson, W. C. & Weiss, W. A. Myc proteins as therapeutic targets. Oncogene29, 1249–1259 (2010). 10.1038/onc.2009.512
+Weiss, W. A., Aldape, K., Mohapatra, G., Feuerstein, B. G. & Bishop, J. M. Targeted expression of MYCN causes neuroblastoma in transgenic mice. EMBO J16, 2985–2995 (1997). 10.1093/emboj/16.11.2985
+Zhu, S.et al.Activated ALK collaborates with MYCN in neuroblastoma pathogenesis. Cancer Cell21, 362–373 (2012). 10.1016/j.ccr.2012.02.010
+Tao, L.et al.MYCN-driven fatty acid uptake is a metabolic vulnerability in neuroblastoma. Nat Commun13, 3728 (2022). 10.1038/s41467-022-31331-2
+Khan, A.et al.Targeting metabolic activity in high-risk neuroblastoma through Monocarboxylate Transporter 1 (MCT1) inhibition. Oncogene39, 3555–3570 (2020). 10.1038/s41388-020-1235-2
+Gamble, L. D.et al.Inhibition of polyamine synthesis and uptake reduces tumor progression and prolongs survival in mouse models of neuroblastoma. Sci Transl Med11 (2019). 10.1126/scitranslmed.aau1099
+Xia, Y.et al.Metabolic Reprogramming by MYCN Confers Dependence on the Serine-Glycine-One-Carbon Biosynthetic Pathway. Cancer Res79, 3837–3850 (2019). 10.1158/0008-5472.CAN-18-3541
+Wang, T.et al.MYCN drives glutaminolysis in neuroblastoma and confers sensitivity to an ROS augmenting agent. Cell Death Dis9, (2018). 10.1038/s41419-018-0295-5
+Bansal, M., Gupta, A. & Ding, H. F. MYCN and Metabolic Reprogramming in Neuroblastoma. Cancers (Basel) 14 (2022). 10.3390/cancers14174113
+Olsen, T. K.et al.DHODH is an independent prognostic marker and potent therapeutic target in neuroblastoma. JCI Insight7 (2022). 10.1172/jci.insight.153836
+Alborzinia, H.et al.MYCN mediates cysteine addiction and sensitizes neuroblastoma to ferroptosis. Nat Cancer3, 471–485 (2022). 10.1038/s43018-022-00355-4
+Guo, X., Chen, Q. R., Song, Y. K., Wei, J. S. & Khan, J. Exon array analysis reveals neuroblastoma tumors have distinct alternative splicing patterns according to stage and MYCN amplification status. BMC Med Genomics4, (2011). 10.1186/1755-8794-4-35
+Shi, Y.et al.Aberrant splicing in neuroblastoma generates RNA-fusion transcripts and provides vulnerability to spliceosome inhibitors. Nucleic Acids Res (2021). 10.1093/nar/gkab054
+Shi, Y.et al.Aberrant splicing in neuroblastoma generates RNA-fusion transcripts and provides vulnerability to spliceosome inhibitors. Nucleic Acids Res49, 2509–2521 (2021). 10.1093/nar/gkab054
+Singh, S.et al.Targeting the spliceosome through RBM39 degradation results in exceptional responses in high-risk neuroblastoma models. Sci Adv7, eabj5405 (2021). 10.1126/sciadv.abj5405
+Nijhuis, A.et al.Indisulam targets RNA splicing and metabolism to serve as a therapeutic strategy for high-risk neuroblastoma. Nat Commun13, 1380 (2022). 10.1038/s41467-022-28907-3
+Pugh, T. J.et al.The genetic landscape of high-risk neuroblastoma. Nat Genet45, 279–284 (2013). 10.1038/ng.2529
+Molenaar, J. J.et al.Sequencing of neuroblastoma identifies chromothripsis and defects in neuritogenesis genes. Nature483, 589–593 (2012). 10.1038/nature10910
+Hagenbuchner, J., Kiechl-Kohlendorfer, U., Obexer, P. & Ausserlechner, M. J. BIRC5/Survivin as a target for glycolysis inhibition in high-stage neuroblastoma. Oncogene35, 2052–2061 (2016). 10.1038/onc.2015.264
+MacArthur, I. C.et al.Prohibitin promotes de-differentiation and is a potential therapeutic target in neuroblastoma. JCI Insight5 (2019). 10.1172/jci.insight.127130
+Milosevic, J.et al.PPM1D Is a Therapeutic Target in Childhood Neural Tumors. Cancers (Basel) 13 (2021). 10.3390/cancers13236042
+Meitinger, F.et al.TRIM37 controls cancer-specific vulnerability to PLK4 inhibition. Nature585, 440–446 (2020). 10.1038/s41586-020-2710-1
+Garcia-Lopez, J.et al.Large 1p36 Deletions Affecting Arid1a Locus Facilitate Mycn-Driven Oncogenesis in Neuroblastoma. Cell Rep30, 454–464 e455 (2020). 10.1016/j.celrep.2019.12.048
+Henrich, K. O.et al.CAMTA1, a 1p36 tumor suppressor candidate, inhibits growth and activates differentiation programs in neuroblastoma cells. Cancer Res71, 3142–3151 (2011). 10.1158/0008-5472.CAN-10-3014
+Liu, Z.et al.CASZ1, a candidate tumor-suppressor gene, suppresses neuroblastoma tumor growth through reprogramming gene expression. Cell Death Differ18, 1174–1183 (2011). 10.1038/cdd.2010.187
+Laut, A. K.et al.CHD5 inhibits metastasis of neuroblastoma. Oncogene41, 622–633 (2022). 10.1038/s41388-021-02081-0
+Higashi, M.et al.Retinoic acid-induced CHD5 upregulation and neuronal differentiation of neuroblastoma. Mol Cancer14, (2015). 10.1186/s12943-015-0425-y
+Fujita, T.et al.CHD5, a tumor suppressor gene deleted from 1p36.31 in neuroblastomas. J Natl Cancer Inst100, 940–949 (2008). 10.1093/jnci/djn176
+Fell, S. M.et al.Neuroblast differentiation during development and in neuroblastoma requires KIF1Bbeta-mediated transport of TRKA. Genes Dev31, 1036–1053 (2017). 10.1101/gad.297077.117
+Li, S.et al.The 1p36 Tumor Suppressor KIF 1Bbeta Is Required for Calcineurin Activation, Controlling Mitochondrial Fission and Apoptosis. Dev Cell36, 164–178 (2016). 10.1016/j.devcel.2015.12.029
+Chen, Z. X.et al.RNA helicase A is a downstream mediator of KIF1Bbeta tumor-suppressor function in neuroblastoma. Cancer Discov4, 434–451 (2014). 10.1158/2159-8290.CD-13-0362
+Cole, K. A.et al.A functional screen identifies miR-34a as a candidate neuroblastoma tumor suppressor gene. Mol Cancer Res6, 735–742 (2008). 10.1158/1541-7786.MCR-07-2102
+Yu, F.et al.RUNX3 interacts with MYCN and facilitates protein degradation in neuroblastoma. Oncogene33, 2601–2609 (2014). 10.1038/onc.2013.221
+Bown, N.et al.Gain of chromosome arm 17q and adverse outcome in patients with neuroblastoma. N Engl J Med340, 1954–1961 (1999). 10.1056/NEJM199906243402504
+Althoff, K.et al.A Cre-conditional MYCN-driven neuroblastoma mouse model as an improved tool for preclinical studies. Oncogene34, 3357–3368 (2015). 10.1038/onc.2014.269
+Bottger, A., Islam, M. S., Chowdhury, R., Schofield, C. J. & Wolf, A. The oxygenase Jmjd6--a case study in conflicting assignments. Biochem J468, 191–202 (2015). 10.1042/BJ20150278
+Chang, B., Chen, Y., Zhao, Y. & Bruick, R. K. JMJD6 is a histone arginine demethylase. Science318, 444–447 (2007). 10.1126/science.1145801
+Webby, C. J.et al.Jmjd6 catalyses lysyl-hydroxylation of U2AF65, a protein associated with RNA splicing. Science325, 90–93 (2009). 10.1126/science.1175865
+Kwok, J., O’Shea, M., Hume, D. A. & Lengeling, A. Jmjd6, a JmjC Dioxygenase with Many Interaction Partners and Pleiotropic Functions. Front Genet8, (2017). 10.3389/fgene.2017.00032
+Vangimalla, S. S., Ganesan, M., Kharbanda, K. K. & Osna, N. A. Bifunctional Enzyme JMJD6 Contributes to Multiple Disease Pathogenesis: New Twist on the Old Story. Biomolecules7 (2017). 10.3390/biom7020041
+Zhou, J.et al.An oncogenic JMJD6-DGAT1 axis tunes the epigenetic regulation of lipid droplet formation in clear cell renal cell carcinoma. Mol Cell82, 3030–3044 e3038 (2022). 10.1016/j.molcel.2022.06.003
+Paschalis, A.et al.JMJD6 Is a Druggable Oxygenase That Regulates AR-V7 Expression in Prostate Cancer. Cancer Res81, 1087–1100 (2021). 10.1158/0008-5472.CAN-20-1807
+Yang, J.et al.Targeting Histone Demethylases in MYC-Driven Neuroblastomas with Ciclopirox. Cancer Res77, 4626–4638 (2017). 10.1158/0008-5472.CAN-16-0826
+Wong, M.et al.JMJD6 is a tumorigenic factor and therapeutic target in neuroblastoma. Nat Commun10, 3319 (2019). 10.1038/s41467-019-11132-w
+Liu, W.et al.Brd4 and JMJD6-associated anti-pause enhancers in regulation of transcriptional pause release. Cell155, 1581–1595 (2013). 10.1016/j.cell.2013.10.056
+Miller, T. E.et al.Transcription elongation factors represent in vivo cancer dependencies in glioblastoma. Nature547, 355–359 (2017). 10.1038/nature23000
+Loven, J.et al.Selective inhibition of tumor oncogenes by disruption of super-enhancers. Cell153, 320–334 (2013). 10.1016/j.cell.2013.03.036
+Chapuy, B.et al.Discovery and characterization of super-enhancer-associated dependencies in diffuse large B cell lymphoma. Cancer Cell24, 777–790 (2013). 10.1016/j.ccr.2013.11.003
+Puissant, A.et al.Targeting MYCN in neuroblastoma by BET bromodomain inhibition. Cancer Discov3, 308–323 (2013). 10.1158/2159-8290.CD-12-0418
+Wyce, A.et al.BET inhibition silences expression of MYCN and BCL2 and induces cytotoxicity in neuroblastoma tumor models. PLoS One8, (2013). 10.1371/journal.pone.0072967
+Meyers, R. M.et al.Computational correction of copy number effect improves specificity of CRISPR-Cas9 essentiality screens in cancer cells. Nat Genet49, 1779–1784 (2017). 10.1038/ng.3984
+Brodeur, G. M. & Bagatell, R. Mechanisms of neuroblastoma regression. Nat Rev Clin Oncol11, 704–713 (2014). 10.1038/nrclinonc.2014.168
+Delmore, J. E.et al.BET bromodomain inhibition as a therapeutic strategy to target c-Myc. Cell146, 904–917 (2011). 10.1016/j.cell.2011.08.017
+Liu, W.et al.Brd4 and JMJD6-associated anti-pause enhancers in regulation of transcriptional pause release. Cell155, 1581–1595 (2013). 10.1016/j.cell.2013.10.056
+Wu, G.et al.Inhibition of SF3B1 by molecules targeting the spliceosome results in massive aberrant exon skipping. RNA24, 1056–1066 (2018). 10.1261/rna.065383.117
+Dang, C. V. MYC, metabolism, cell growth, and tumorigenesis. Cold Spring Harb Perspect Med3 (2013). 10.1101/cshperspect.a014217
+Miller, D. M., Thomas, S. D., Islam, A., Muench, D. & Sedoris, K. c-Myc and cancer metabolism. Clin Cancer Res18, 5546–5553 (2012). 10.1158/1078-0432.CCR-12-0977
+Gordan, J. D., Thompson, C. B. & Simon, M. C. HIF and c-Myc: sibling rivals for control of cancer cell metabolism and proliferation. Cancer Cell12, 108–113 (2007). 10.1016/j.ccr.2007.07.006
+Wise, D. R. & Thompson, C. B. Glutamine addiction: a new therapeutic target in cancer. Trends Biochem Sci35, 427–433 (2010). 10.1016/j.tibs.2010.05.003
+Porter, L. D., Ibrahim, H., Taylor, L. & Curthoys, N. P. Complexity and species variation of the kidney-type glutaminase gene. Physiol Genomics9, 157–166 (2002). 10.1152/physiolgenomics.00017.2002
+Cassago, A.et al.Mitochondrial localization and structure-based phosphate activation mechanism of Glutaminase C with implications for cancer metabolism. Proc Natl Acad Sci U S A109, 1092–1097 (2012). 10.1073/pnas.1112495109
+Wang, J. B.et al.Targeting mitochondrial glutaminase activity inhibits oncogenic transformation. Cancer Cell18, 207–219 (2010). 10.1016/j.ccr.2010.08.009
+Gao, P.et al.c-Myc suppression of miR-23a/b enhances mitochondrial glutaminase expression and glutamine metabolism. Nature458, 762–765 (2009). 10.1038/nature07823
+Wise, D. R.et al.Myc regulates a transcriptional program that stimulates mitochondrial glutaminolysis and leads to glutamine addiction. Proc Natl Acad Sci U S A105, 18782–18787 (2008). 10.1073/pnas.0810199105
+Yuneva, M., Zamboni, N., Oefner, P., Sachidanandam, R. & Lazebnik, Y. Deficiency in glutamine but not glucose induces MYC-dependent apoptosis in human cells. J Cell Biol178, 93–105 (2007). 10.1083/jcb.200703099
+Masamha, C. P.et al.CFIm25 regulates glutaminase alternative terminal exon definition to modulate miR-23 function. RNA22, 830–838 (2016). 10.1261/rna.055939.116
+Yi, J.et al.JMJD6 and U2AF65 co-regulate alternative splicing in both JMJD6 enzymatic activity dependent and independent manner. Nucleic Acids Res45, 3503–3518 (2017). 10.1093/nar/gkw1144
+Li, H.et al.The landscape of cancer cell line metabolism. Nat Med25, 850–860 (2019). 10.1038/s41591-019-0404-8
+David, C. J., Chen, M., Assanah, M., Canoll, P. & Manley, J. L. HnRNP proteins controlled by c-Myc deregulate pyruvate kinase mRNA splicing in cancer. Nature463, 364–368 (2010). 10.1038/nature08697
+Rauch, J.et al.c-Myc regulates RNA splicing of the A-Raf kinase and its activation of the ERK pathway. Cancer Res71, 4664–4674 (2011). 10.1158/0008-5472.CAN-10-4447
+Ge, Y.et al.The splicing factor RBM25 controls MYC activity in acute myeloid leukemia. Nat Commun10, (2019). 10.1038/s41467-018-08076-y
+Seton-Rogers, S. RNA splicing: MYC maintains high-fidelity splicing. Nat Rev Cancer15, (2015). 10.1038/nrc3977
+Matera, A. G. & Wang, Z. A day in the life of the spliceosome. Nat Rev Mol Cell Biol15, 108–121 (2014). 10.1038/nrm3742
+Wahl, M. C., Will, C. L. & Luhrmann, R. The spliceosome: design principles of a dynamic RNP machine. Cell136, 701–718 (2009). 10.1016/j.cell.2009.02.009
+Pan, Q., Shai, O., Lee, L. J., Frey, B. J. & Blencowe, B. J. Deep surveying of alternative splicing complexity in the human transcriptome by high-throughput sequencing. Nat Genet40, 1413–1415 (2008). 10.1038/ng.259
+Wang, E. T.et al.Alternative isoform regulation in human tissue transcriptomes. Nature456, 470–476 (2008). 10.1038/nature07509
+Yang, X.et al.Widespread Expansion of Protein Interaction Capabilities by Alternative Splicing. Cell164, 805–817 (2016). 10.1016/j.cell.2016.01.029
+Liu, Y.et al.Impact of Alternative Splicing on the Human Proteome. Cell Rep20, 1229–1241 (2017). 10.1016/j.celrep.2017.07.025
+Daemen, A.et al.Pan-Cancer Metabolic Signature Predicts Co-Dependency on Glutaminase and De Novo Glutathione Synthesis Linked to a High-Mesenchymal Cell State. Cell Metab28, 383–399 e389 (2018). 10.1016/j.cmet.2018.06.003
+Redis, R. S.et al.Allele-Specific Reprogramming of Cancer Metabolism by the Long Non-coding RNA CCAT2. Mol Cell61, 520–534 (2016). 10.1016/j.molcel.2016.01.015
+Gong, C.et al.Sequential inverse dysregulation of the RNA helicases DDX3X and DDX3Y facilitates MYC-driven lymphomagenesis. Mol Cell (2021). 10.1016/j.molcel.2021.07.041
+Bosse, K. R. & Maris, J. M. Advances in the translational genomics of neuroblastoma: From improving risk stratification and revealing novel biology to identifying actionable genomic alterations. Cancer (2015). 10.1002/cncr.29706
+Pinto, N. R.et al.Advances in Risk Classification and Treatment Strategies for Neuroblastoma. J Clin Oncol33, 3008–3017 (2015). 10.1200/JCO.2014.59.4648
+Brodeur, G. M. Neuroblastoma: biological insights into a clinical enigma. Nat Rev Cancer3, 203–216 (2003). 10.1038/nrc1014
+Maris, J. M., Hogarty, M. D., Bagatell, R. & Cohn, S. L. Neuroblastoma. Lancet369, 2106–2120 (2007). 10.1016/S0140-6736(07)60983-0
+Cohn, S. L.et al.The International Neuroblastoma Risk Group (INRG) classification system: an INRG Task Force report. J Clin Oncol27, 289–297 (2009). 10.1200/jco.2008.16.6785
+Suh, E.et al.Late mortality and chronic health conditions in long-term survivors of early-adolescent and young adult cancers: a retrospective cohort analysis from the Childhood Cancer Survivor Study. Lancet Oncol21, 421–435 (2020). 10.1016/S1470-2045(19)30800-9
+Nathan, P. C.et al.Health-related quality of life in adult survivors of childhood Wilms tumor or neuroblastoma: A report from the childhood cancer survivor study. Pediatr Blood Cancer49, 704–715 (2007). 10.1002/pbc.20949
+Brady, S. W.et al.Pan-neuroblastoma analysis reveals age- and signature-associated driver alterations. Nat Commun11, 5183 (2020). 10.1038/s41467-020-18987-4
+Han, T.et al.Anticancer sulfonamides target splicing by inducing RBM39 degradation via recruitment to DCAF15. Science356 (2017). 10.1126/science.aal3755
+Xu, P., Duong, D. M. & Peng, J. Systematical optimization of reverse-phase chromatography for shotgun proteomics. J Proteome Res8, 3944–3950 (2009). 10.1021/pr900251d
+Bai, B.et al.U1 small nuclear ribonucleoprotein complex and RNA splicing alterations in Alzheimer’s disease. Proc Natl Acad Sci U S A110, 16562–16567 (2013). 10.1073/pnas.1310249110
+Dobin, A.et al.STAR: ultrafast universal RNA-seq aligner. Bioinformatics29, 15–21 (2013). 10.1093/bioinformatics/bts635
+Shannon, P.et al.Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res13, 2498–2504 (2003). 10.1101/gr.1239303
+
+
+
diff --git a/test/2023.08.03.551564/fix-refs/2023.08.03.551564.xml b/test/2023.08.03.551564/fix-refs/2023.08.03.551564.xml
new file mode 100644
index 00000000..70e734d0
--- /dev/null
+++ b/test/2023.08.03.551564/fix-refs/2023.08.03.551564.xml
@@ -0,0 +1,511 @@
+
+
+
+
+
+BIORXIV
+
+bioRxiv
+bioRxiv
+
+
+Cold Spring Harbor Laboratory
+
+
+
+10.1101/2023.08.03.551564
+1.1
+
+
+Regular Article
+
+
+New Results
+
+
+Neuroscience
+
+
+
+Restoration of locomotor function following stimulation of the A13 region in Parkinson’s mouse models
+
+
+
+http://orcid.org/0000-0003-0628-3104
+KimLinda H
+1
+2
+*
+
+
+http://orcid.org/0000-0002-6067-4272
+LognonAdam
+1
+2
+*
+
+
+SharmaSandeep
+1
+3
+
+
+http://orcid.org/0000-0002-2856-919X
+TranMichelle A.
+3
+
+
+http://orcid.org/0000-0001-6118-813X
+ChomiakTaylor
+1
+4
+
+
+TamStephanie
+3
+
+
+McPhersonClaire
+3
+
+
+http://orcid.org/0000-0002-8361-1733
+EatonShane E. A.
+1
+3
+
+
+http://orcid.org/0000-0002-8656-2690
+KissZelma H. T.
+1
+2
+4
+
+
+http://orcid.org/0000-0002-1234-5415
+WhelanPatrick J.
+1
+3
+#
+
+Hotchkiss Brain Institute, University of Calgary, Calgary, AB, Canada, T2N4N1,
+Department of Neuroscience, University of Calgary, Calgary, AB, Canada, T2N 4N1,
+Faculty of Veterinary Medicine, University of Calgary, Calgary, AB, Canada, T2N4N1,
+Department of Clinical Neurosciences, University of Calgary, Calgary, AB, Canada, T2N 4N1
+
+
+Corresponding author Patrick J. Whelan HMRB 168, 3330 Hospital Drive NW, University of Calgary Calgary, AB T2N 4N1 Email: whelan@ucalgary.ca
+
Parkinson’s disease (PD) is characterized by extensive motor and non-motor dysfunction, including gait disturbance, which is difficult to treat effectively. This study explores the therapeutic potential of targeting the A13 region, a dopamine-containing area of the medial zona incerta (mZI) that has shown relative preservation in PD models. The A13 is identified to project to the mesencephalic locomotor region (MLR), with a subpopulation of cells displaying activity correlating to movement speed, suggesting its potential involvement in locomotor function. We show that photoactivation of this region can alleviate bradykinesia and akinetic symptoms in a mouse model of PD, revealing the presence of preserved parallel motor pathways for movement. We identified areas of preservation and plasticity within the mZI connectome using whole-brain imaging. Our findings suggest a global remodeling of afferent and efferent projections of the A13 region, highlighting the zona incerta’s role as a crucial hub for the rapid selection of motor function. Despite endogenous compensatory mechanisms proving insufficient to overcome locomotor deficits in PD, our data demonstrate that photostimulation of the A13 region effectively restores locomotor activity. The study unveils the significant pro-locomotor effects of the A13 region and suggests its promising potential as a therapeutic target for PD-related gait dysfunction.
+
+SIGNIFICANCE STATEMENT
+
This work examines the function of the A13 nucleus in locomotion, an area with direct connectivity to locomotor regions in the brainstem. Our work shows that A13 stimulation can restore locomotor function and improve bradykinesia symptoms in a PD mouse model.
Parkinson’s disease (PD) is a complex condition affecting many facets of motor and non-motor functions, including visual, olfactory, memory and executive functions (Cenci and Björklund, 2020). Due to the widespread features of PD, focusing on changes within a single pathway cannot account for all symptoms. Gait disturbance is one of the hardest to treat; pharmacological, deep brain stimulation (DBS) and physical therapies lead to only partial improvements (Nonnekes et al., 2020, 2015). While the subthalamic nucleus (STN) and globus pallidus (GPi) are common DBS targets for PD, alternative targets such as pedunculopontine nucleus (PPN) and the zona incerta (ZI) have been proposed with mixed results in improving postural and/or gait dysfunctions (Caire et al., 2013; Ferraye et al., 2010; Gut and Winn, 2015; Hamani et al., 2011; Moro et al., 2010; Nonnekes et al., 2015; Okun and Foote, 2010; Ossowska, 2019; Stefani et al., 2007; Thevathasan et al., 2018). Part of the issue with targeting the ZI with DBS strategies is the relative lack of knowledge regarding its downstream anatomical and functional connectivity with motor centres. Recent work with photoactivation of subpopulations of PPN neurons in PD models shows promise for similar ZI-focused strategies (Masini and Kiehn, 2022).
+
The ZI is recognized as an integrative hub, with roles in regulating sensory inflow, arousal, motor function, and conveying motivational states (Mitrofanis, 2005; Wang et al., 2020). As such, it is well placed to be involved in PD and has seen increased clinical and preclinical research over the last two decades (Blomstedt et al., 2018; Ossowska, 2019; Plaha et al., 2008). However, little attention has been placed on the medial zona incerta (mZI), particularly the A13, the only dopamine-containing region of the rostral ZI (Bolton et al., 2015; Kim et al., 2017; Sharma et al., 2018). Recent research in primates and mice (Peoples et al., 2012; Roostalu et al., 2019; Shaw et al., 2010) indicates that the A13 is preserved in 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-based PD models.
+
Recently, we discovered that the A13 located within the ZI projects to two areas of the mesencephalic locomotor region (MLR), the PPN and cuneiform nucleus (CnF)(Sharma et al., 2018), suggesting a role for A13 in locomotor function. Indeed, in vivo photometry recordings from calcium/calmodulin-dependent protein kinase IIα (CaMKIIα) populations in the rostral ZI, which includes the A13 nucleus, show a subpopulation of cells whose activity correlates with movement speed (Li et al., 2021). Since this region projects to the MLR, it is a potential parallel motor pathway to target for gait improvement. Photoactivation of glutamatergic MLR neurons alleviates motor deficits in the 6-OHDA mouse model (Fougère et al., 2021; Masini and Kiehn, 2022). Phenomena such as kinesia paradoxa (Glickstein and Stein, 1991) in PD patients support the existence of preserved parallel motor pathways that can be engaged in particular circumstances to produce normal movement.
+
Further evidence supporting the importance of parallel motor pathways in PD includes those reporting functional alterations in A13 (Hoffman et al., 1997; Périer et al., 2000). Nigrostriatal lesions affect A13 cellular function and lead to anatomical remodeling in monoaminergic brain regions (Braak et al., 2003; Kish et al., 2008; Lim et al., 2009; Perez-Lloret and Barrantes, 2016; Roostalu et al., 2019; Scatton et al., 1983; Zweig et al., 1989). The A13 connectome encompasses the cerebral cortex (Mitrofanis and Mikuletic, 1999), central nucleus of the amygdala (Eaton et al., 1994), thalamic paraventricular nucleus (Li et al., 2014), thalamic reuniens (Sita et al., 2007; Venkataraman et al., 2021), MLR(Sharma et al., 2018), superior colliculus (SC) (Bolton et al., 2015), and dorsolateral periaqueductal grey (PAG) (Messanvi et al., 2013; Sita et al., 2007), making the A13 an important hub for goal-directed locomotion (Choi and McNally, 2017; Eaton et al., 1994; Messanvi et al., 2013; Mok and Mogenson, 1986; Moriya et al., 2020; Ogundele et al., 2017; Manjit K. Sanghera et al., 1991; M. K. Sanghera et al., 1991; Sita et al., 2007; Venkataraman et al., 2021).
+
Based on the role of the A13 in gait and, specifically, as a possible target to improve gait in PD, we investigated the therapeutic potential of photoactivating a tightly circumscribed region targeting a small region containing mainly the A13 and a small area of the mZI, which we term A13 region throughout the manuscript. We identified areas of preservation and plasticity within the mZI connectome using whole-brain imaging techniques. Photoactivation of the A13 region rescued bradykinetic and akinetic symptoms in a mouse model of 6-hydroxydopamine (6-OHDA) mediated unilateral nigrostriatal degeneration. Because the zona incerta is a hub for the rapid selection of motor function, we then mapped the input and output patterns of the region. We found evidence of a global remodeling of afferent and efferent projections of the A13 region. While endogenous compensatory mechanisms from the remaining but remodelled A13 region connectome were inadequate in overcoming locomotor deficits observed in 6-OHDA mice, photostimulation of the A13 region restored locomotor activity. These data demonstrate that the A13 region produces powerful pro-locomotor effects in normal and PD mouse models. Moreover, PD-related bradykinesia is ameliorated with A13 region photoactivation in the presence of remodelling of the A13 region connectome. Some of these data have been published in abstract form (L. Kim et al., 2021).
+
+
+RESULTS
+
+Unilateral 6-OHDA mouse model has robust motor deficits
+
The overall experimental design is illustrated in Figure 1A, along with a schematic in Figure 1B showing injections of 6-OHDA in the medial forebrain bundle and AAVDJ-CaMKIIα-ChR2 virus into the medial zona incerta (mZI). We confirmed substantia nigra pars compacta (SNc) degeneration in a well-validated unilateral 6-OHDA-mediated Parkinsonian mouse model (Thiele et al., 2012). The percentage of tyrosine hydroxylase (TH+) cell loss normalized to the intra-animal contralesional side was quantified. 6-OHDA produced a significant lesion that decreased TH+ neuronal SNc populations. As previously reported (Boix et al., 2015), the SNc ipilesional to the 6-OHDA injection (n = 10) showed major ablation of the TH+ neurons compared to sham animals (Figure 1C and D: n = 11).
+
+
+
Experimental design and confirmation of unilateral TH<sup>+</sup> depletion in the SNc via 6-OHDA lesion.
+
(A) Illustration of experimental timeline. (B) Dual ipsilateral stereotaxic injection into the MFB and A13 region. (C) TH+ cells in SNc of sham (top) compared to 6-OHDA injected mouse (bottom). Magnified areas outlined by yellow squares are shown on the right. (D) Unilateral injection of 6-OHDA (6-OHDA ChR2: n = 5, 6-OHDA eYFP: n = 5) into the MFB resulted in greater percentage of TH+ loss compared to sham in the SNc (sham ChR2: n = 7, sham eYFP: n = 5, three-way MM ANOVAs), regardless of virus type (F1,18 = 104.4, p < .001). ***p < .001. Error bars indicate SEMs.
+
+
+
+
+A13 region photoactivation generates pro-locomotor behaviors in the open field
+
6-OHDA lesions are characterized as generating bradykinetic and akinetic phenotypes in the open field (Li et al., 2022; Magno et al., 2019; Masini and Kiehn, 2022; Sanders and Jaeger, 2016). To understand the impact of A13 region photoactivation on locomotion in sham and PD model mice, on-target localization of ferrule above the A13 region, centered on the mZI, along with YFP reporter expression, was confirmed (Figure 2) in mice given sham or 6-OHDA injections. Corroborating the post hoc targeting, we found evidence for c-Fos in neurons within the A13 region in photostimulated ChR2 mice (Figure 2). Before post hoc analysis, mice were monitored in the open field test (OFT), where the effects of the 6-OHDA lesion were apparent, with 6-OHDA lesioned animals demonstrating far less movement, fewer bouts of locomotion, and less time engaging in locomotion in the OFT (Figure 2A-E). Notably, photoactivation of the A13 region often generated dramatic effects, with mice showing a distinct increase in locomotor behavior (Figure 3A, Movie S1 & Movie S2). Both sham and 6-OHDA ChR2 mice showed a significant increase in locomotor distance travelled during periods of photoactivation (Figure 3B, p = 0.005). One sham animal showed grooming behavior on stimulation and was excluded from the analysis.
+
+
+
<italic>Post hoc</italic> c-Fos expression and targeting of the mZI and A13.
+
(A) Diagram showing the A13 DAergic nucleus in dark magenta encapsulated by the ZI in light magenta. The fibre optic tip is outlined in red. Atlas image adapted from the Allen Brain Atlas (Goldowitz, 2010). (B) Tissue images were obtained from 6-OHDA ChR2 animals around bregma −1.22 mm, and (C) a 6-OHDA eYFP animal more caudally around bregma −1.46 mm. Images show the distribution of DAPI (blue), eYFP (green), c-Fos (yellow), and TH (magenta). Landmarks are outlined in white (3V: third ventricle; mtt: mammillothalamic tract), and the optic cannula tip is shown in red. Higher magnification images of the A13 DAergic nucleus are outlined by the yellow boxes in a 6-OHDA ChR2 animal (D) and a 6-OHDA eYFP animal (E). Scale bars are set to 350 μm. Images show isolated channels in the top rows of the respective groups: eYFP (i), TH (ii), and c-Fos (iii). Merged channels for eYFP and c-Fos (iv), TH and c-Fos (v), and a merge of all four channels (vi) are presented in the bottom rows of their respective groups. White arrowheads in the merged images highlight overlap in merged markers. Red arrows show triple colocalization of eYFP, c-Fos and TH. (Dvi) contains a magnified example of triple-labelled neurons, as highlighted in the yellow box. Scale bars are set to 50 μm.
+
+
+
+
+
Ispilesional photoactivation of the A13 region in a unilateral 6-OHDA mouse model rescues motor deficits.
+
(A) Schematic of open field experiment design and example traces for open field testing (1 min) with unilateral photoactivation of the A13 region. (B-E) Effects of photoactivation on open field metrics for sham eYFP (n = 5), sham ChR2 (n = 6), 6-OHDA eYFP (n = 5), and 6-OHDA ChR2 (n = 5) groups (three-way MM ANOVAs, post hoc Bonferroni pairwise). Photoactivation increased in the ChR2 groups: (B) distance travelled (ChR2 vs. eYFP: p = 0.005), (C) locomotor bouts (ChR2 vs. eYFP: p = 0.005), (D) duration of locomotion in the open field (ChR2 vs. eYFP: p = 0.005), and (E) animal movement speed (ChR2 vs. eYFP: p < 0.001). (F-I) Group averaged instantaneous velocity graphs showing no increase in a sham eYFP (F) or 6-OHDA eYFP mouse (H), with increases in velocity during stimulation in a sham ChR2 (G) and 6-OHDA ChR2 (I) mouse. (J) The graph presents animal rotational bias using the turn angle sum. There was a significant increase in 6-OHDA ChR2 rotational bias during A13 region photoactivation (6-OHDA ChR2 vs. 6-OHDA eYFP: p < 0.001). (K) Diagram depicting the pole test. A mouse is placed on a vertical pole facing upwards. The time for release is taken as the experimenter removes their hand from the animal’s tail. (L, M) Graphs showing the response of animals to photoactivation of the A13 region while performing the pole test. (L) A13 region photoactivation also led to shorter total descent time in ChR2 compared to eYFP mice (ChR2 vs. eYFP: p = 0.004), and (M) 6-OHDA ChR2 mice showed a greater reduction in descent time compared to sham ChR2 (6-OHDA ChR2 vs. sham ChR2: p = 0.012; 6-OHDA ChR2: n = 5; sham ChR2: n = 7). ***p < .001, **p < .01, *p < .05. Bonferroni’s post hoc comparisons between 6-OHDA ChR2 and sham eYFP, sham ChR2, and 6-OHDA eYFP at stim time point as a, b, and c respectively. Error bars indicate SEMs.
+
+
+
We tested whether photoactivation led to a single bout of locomotion or if there was an overall increase in bouts, signifying that animals could repeatedly initiate locomotion following photoactivation. Mice in the ChR2 groups demonstrated an increase in the number of locomotion bouts with photoactivation, indicating a greater ability to start locomotion from rest, and that photoactivation was not eliciting a single prolonged bout (Figure 3C, p = 0.005). When we examined each bout of locomotion, photoactivation increased the total duration of locomotion (Figure 3D, p = 0.005). There was a refractory decrease in the distance travelled by the sham ChR2 animal group (Figure S1A), which was not evident for the 6-OHDA cohort (Figure S1B). To control for this, we compared the pre-timepoints to the baseline one-minute averages to ensure that the animal locomotion distance travelled returned to a stable state before stimulation was reapplied (Figure S1C, p = 0.783).
+
Next, we examined the reliability of photoactivation to initiate locomotion. The percentage of trials with at least one bout of locomotion was compared for the pre-and stim time points. 6-OHDA ChR2 animals showed a reliable pro-locomotion phenotype with A13 region photoactivation (Figure S2A: p = 0.042). As was expected in the control 6-OHDA eYFP group, there was no effect of photoactivation on the probability of engaging in locomotion (p = 0.713).
+
Animal movement speed also factors into the total distance travelled measure and can be discussed in regard to a bradykinetic phenotype in 6-OHDA lesioned mice (Magno et al., 2019; Masini and Kiehn, 2022; Sanders and Jaeger, 2016). Using instantaneous animal movement speeds that exceeded 2 cm/s as per Masini & Kiehn (2022), we plotted instantaneous speed (Figure 3F-I) and analyzed one-minute bins (Figure 3E). As was expected, 6-OHDA lesioned animals had lower movement speeds than sham control animals (p < 0.001). One animal from the 6-OHDA eYFP group was excluded because it did not meet the speed threshold during recording. Both the 6-OHDA ChR2 and sham ChR2 groups displayed increases in average speed during photostimulation (Figure 3E, p < 0.001). When we examined the time taken to initiate locomotion, there was no significant difference between sham or 6-OHDA ChR2 groups (Figure S2B).
+
+
+Photoactivation of the A13 region increases ipsilesional turning in the open field test
+
Unilateral 6-OHDA lesions drive asymmetric rotational bias (Boix et al., 2015; Li et al., 2022; Magno et al., 2019; Thiele et al., 2012). We were interested in whether this persisted with stimulation and noted upon observation of photoactivation that animals appeared to have increased ipsilesional rotation. As observed, 6-OHDA ChR2 animals had an increase in turn angle sum (TAS), indicating an increase in their rotational bias with photoactivation in the ipsilesional direction (Figure 3J, p < 0.001). As expected, 6-OHDA eYFP animals showed consistent rotational bias throughout time. The rotational bias of sham ChR2 was also compared to determine whether the increased rotational bias was due to photoactivation or the interaction of photoactivation and the lesion. The sham ChR2 group showed no significant change in TAS with photoactivation (Figure 3J, p > 0.05). Next, we examined whether the increased turning angle sum in the 6-OHDA ChR2 group was observed during periods of locomotion. When the TAS was calculated only during periods of locomotion, the rotational bias in the animal orientation in the 6-OHDA ChR2 animals was not observed (p = 0.286).
+
+
+Skilled vertical locomotion is improved in the pole test with photoactivation of the A13 region
+
The pole test is a classic 6-OHDA behavioral paradigm (Figure 3K) that involves skilled locomotor abilities for an animal to turn and descend a vertical pole (Matsuura et al., 1997; Ogawa et al., 1985). Improvements in function can be inferred if the time taken to complete the test decreases (Matsuura et al., 1997; Ogawa et al., 1985). 6-OHDA mice demonstrated significantly greater descent times than sham mice (p < 0.001). Photoactivation of the A13 region reduced descent times for both 6-OHDA and sham groups on the pole test (Figure 3L, p = 0.004, Movie S3). Neither of the eYFP groups showed any changes in the time to complete the pole test.
+
To further understand the effects of photoactivation on the ability of mice to descend the pole, the time taken for mice to descend after turning was analyzed to remove any influence of animals spending time investigating their environment on the top of the pole. While all groups showed reduced total pole test descent time with photoactivation, considering just the time to descend from turn alone, there was a larger improvement with A13 region photoactivation in the 6-OHDA ChR2 mice compared to sham ChR2 mice (Figure 3M: p = 0.012). These results indicate that photoactivation has the effect of reducing bradykinesia by improving the ability of mice to descend the pole during the PT.
+
+
+Dopaminergic Cells in the A13 region are preserved in the unilateral 6-OHDA mouse model
+
While photoactivation of the A13 region promoted locomotor activity in both sham and 6-OHDA mice, there were differences in speed and directional bias. We hypothesized that this may be due to changes in the A13 region connectome since there is evidence of changes in firing and metabolic activity in the region (Périer et al., 2000). Therefore, we utilized whole brain imaging approaches (Hansen et al., 2020; Zhan et al., 2021) to examine changes in the connectome following 6-OHDA lesions of the nigrostriatal region.
+
Using whole brain imaging, as expected, TH+ cells in SNc were more vulnerable to the 6-OHDA neurotoxin than the ventral tegmental area (VTA) and A13 (Figure S3A-F). 6-OHDA-treated mice showed a significantly greater percentage of TH+ cell loss in SNc compared to the VTA and A13 (VTA vs. SNc: p = 0.003; A13 vs. SNc: p = 0.005). In contrast, sham animals showed no significant difference in TH+ cell loss across SNc, VTA and A13 (Figure S3G, p > 0.05). Thus, similar to that observed in the human brain of Parkinsonian patients (Matzuk and Saper, 1985), there is a remarkable preservation of dopaminergic cells in the A13 after nigrostriatal degeneration in the 6-OHDA mouse model of PD.
+
+
+Large-scale changes in the A13 region connectome following 6-OHDA-mediated unilateral nigrostriatal degeneration
+
Although photoactivation had benefits in restoring speed in 6-OHDA mice (Figure 3), circling behavior was increased, suggesting additional changes that may reflect connectome alterations. We examined the changes in the input and output of the A13 by co-injecting anterograde (AAV8-CamKII-mCherry) and retrograde AAV (AAVrg-CAG-GFP) tracers into the A13 nucleus (Keith B. J. Franklin and Paxinos, 2008). The injection core and spread were determined in the rostrocaudal direction from the injection site (Figure S4). To examine whether unilateral nigrostriatal degeneration resulted in changes in the organization of inputs and outputs from the A13, we first visualized interregional correlations of afferent and efferent proportions for each condition using correlation matrices (Fig 4A and B; 251 regions in a pairwise manner). Correlation matrices were organized using the hierarchical anatomical groups from the Allen Brain Atlas (Figure 4C). To minimize the influence of experimental variation on the total labeling of neurons and fibers, the afferent cell counts or efferent fiber areas in each brain region were divided by the total number found in a brain to obtain the proportion of total inputs and outputs. The data were normalized to a log10 value to reduce variability and bring brain regions with high and low proportions of cells and fibers to a similar scale (Kimbrough et al., 2020). Comparing the afferent and efferent proportions in a pairwise manner between mice showed good consistency with an average correlation of 0.91 ± 0.02 (Spearman’s correlation, Figure S5).
+
+
+
Unilateral nigrostriatal degeneration leads to large-scale changes in the organization of the A13 region afferent and efferent distributions across the neuraxis.
+
We used correlation matrices to summarize any observable patterns in the distribution patterns of inputs and outputs of the A13 region. A correlation matrix was calculated by correlating the proportion of input from one brain region to another in a pairwise manner across 251 brain regions delimited by registration with Allen Brain Atlas. If two brain regions among mice (eg. brain regions A and B) contribute a similar input, they are highly correlated (A). Using the color legend showing various correlation strengths, the intersecting box in the matrix in this example will be colored dark red (B). If no relationship is found between contributions from two brain regions, the intersecting box will be colored yellow. If the contribution from one brain region was negatively correlated with another brain region among mice, then the intersecting box will be colored blue. The afferent distribution pattern in the sham displayed a higher level of inter-regional correlation between brain regions (C) than 6-OHDA injected mice (D). Indeed, two distinct bands of anti-correlated afferent regions were identified in the 6-OHDA injected mice (see black boxes in D). These two bands arose from the cortical plate subregions (motor, sensory, visual, and prefrontal), and striatal and pallidal subregions showing distinct inputs compared to the rest of the neuraxis. In contrast, the projection patterns of A13 efferents displayed a higher level of inter-regional correlation between brain regions following a unilateral nigrostriatal degeneration (F) compared to sham (E). In sham, proportions of A13-cortical/ striatal efferents were negatively correlated to A13-pallidal/ thalamic/ hypothalamic/ midbrain efferents (see black boxes in E). However, these distinct projection patterns disappeared following nigrostriatal degeneration, suggesting A13 efferent distributions becoming more distributed across the neuraxis.
+
+
+
We observed changes in projection patterns between the sham and the 6-OHDA group. A correlation matrix was used to quantify the relationship of input from brain regions in a pairwise manner through the neuraxis. Overall, afferents onto the A13 in sham animals displayed a higher interregional correlation between brain regions than 6-OHDA-injected mice. Specifically, correlation coefficients of 0.10, 0.30, and 0.50 or larger represent weak, moderate, and strong correlations, respectively (Cohen, 1988). Compared to sham (Figure 4C), in 6-OHDA injected mice (Figure 4D), afferent contributions from two clusters of brain subregions became more dissimilar (anti-correlated) to the rest of the neuraxis: 1) cortical plate subregions (motor, sensory, visual, and prefrontal), and 2) striatal, and pallidal subregions compared to sham (boxed blue areas in Figure 4D). These data suggest that afferents from several regions showed a coordinated reduction in afferent density onto the A13, except contributions from cortical plate (motor, sensory, visual, and prefrontal), striatal, and pallidal subregions that were positively correlated. In other words, contributions from the cortical plate (motor, sensory, visual, and prefrontal), striatal and pallidal subregions are positively correlated, but compared to the rest of the brain, they are anti-correlated. This suggests a greater afferent input onto A13 from the cortical plate (motor, sensory, visual, and prefrontal), striatal and pallidal subregions than other regions after 6-OHDA lesions.
+
In marked contrast, the projection patterns of A13 efferents exhibited a higher level of interregional correlation between brain regions following a unilateral nigrostriatal degeneration compared to sham. In the sham condition, the A13 connectome is biased towards cortical and striatal regions compared to pallidal, thalamic, hypothalamic, midbrain efferents. This is shown by a broad negative correlation between these two large groups (Figure 4E). However, these broad anti-correlations disappear following nigrostriatal degeneration (Figure 4F). These data indicate that the A13 efferent connectome is less refined following nigrostriatal degeneration.
+
+
+Differential remodeling of A13 region connectome ipsi- and contra-lesion following 6-OHDA-mediated nigrostriatal degeneration
+
The distributions of the A13 connectome in sham animals served as the basis for an in-depth comparison of the preservation and plasticity of A13 afferents and efferents in 6-OHDA mouse models (Figure 5). We observed a global remodeling of A13 afferents and efferents following unilateral nigrostriatal degeneration (Figure 5A, B; E, F) that was differentially expressed across the neuraxis (Figure 4D, H). The ipsilesional side showed more downregulated areas (Figure 5D: see also example traces). These downregulated areas were focused within the cortical plate and cortical subplate regions. 6-OHDA injections also downregulated A13 afferent densities from the striatum, pallidum, thalamus and medulla. While 6-OHDA mainly downregulated ipsilesional A13 afferent densities, the hypothalamus (including ZI), midbrain, pons, and cerebellum had increased A13 afferent densities.
+
+
+
Differential remodeling of A13 region connectome following a unilateral nigrostriatal degeneration.
+
The distributions of the A13 connectome in sham served as a basis for an in-depth comparison against 6-OHDA mouse models. Example registered slices (using WholeBrain software 64 with light-sheet data, 2X objective, 4X optical zoom) at rostral areas show changes in sham (A) afferents compared to 6-OHDA lesioned animals (B). Graph showing major brain regions contributing afferents to A13 in sham mice (C). The graph illustrates the change in the proportion of afferents in 6-OHDA compared to sham mice (D). Representative registered slices showing sham proportions of efferents in sham (E) compared to 6-OHDA mice (F). The magnified black box section displays an example of mCherry+ fibers (left) segmented using Ilastik and ImageJ software (right). Graph showing major brain regions receiving efferents from the A13 in sham mice (G). The graph illustrates the change in the proportion of efferents in 6-OHDA compared to sham mice (H). Error bars represent SEMs. Anterograde and retrograde viruses were injected into the ipsilesional A13 (see methods). Abbreviations from Allen Brain. Atlas: CTXpl (cortical plate), CTXsp (cortical subplate), STR (striatum), PAL (palladium), TH (thalamus), HYP (hypothalamus), P (pons), MB (midbrain), MY (Medulla), and CB (cerebellum).
+
+
+
The intact, contralesional side showed more upregulated regions across the neuraxis, suggesting compensatory upregulation in the unilateral 6-OHDA model (Figure 5D). The cortical plate, striatum, and cortical subplate were the top three upregulated contralesional regions. In contrast, contralesional A13 afferent densities from the pallidum and thalamus were spared and upregulated. Thus, compensatory upregulation of A13 afferent density from these regions appeared lateralized from the intact, contralesional side. Furthermore, bilateral compensatory upregulation of A13 afferents was observed from the hypothalamus (including ZI), midbrain and pons.
+
The A13 efferents were more downregulated on the ipsilesional side (Figure 5H). However, the downregulation was focused within the isocortical, striatal and cortical subplate regions. Remodeling on the contralesional efferent projection patterns closely followed the changes seen with the afferents, except for projections onto the thalamic and midbrain regions. The A13 efferents onto thalamic regions were bilaterally upregulated. Also, the A13-midbrain efferents were upregulated ipsilesionally.
+
+
+
+DISCUSSION
+
Our work demonstrates robust pro-locomotor effects induced by photoactivation of the A13 region in lesioned and sham mice. Photoactivation during the OFT increased locomotion distance travelled, the duration of locomotion, and speed in both sham and 6-OHDA mice. Uniquely, the 6-OHDA group had increases in the number of locomotor bouts which resembled the normal number of bouts observed in healthy mice at baseline. Bradykinesia in 6-OHDA mice was substantially improved following photoactivation. We found extensive input and output connectivity of the A13, which was remodeled following nigrostriatal lesions. Afferent input patterns displayed a marked reduction in interregional correlation across brain regions in 6-OHDA mice, while efferent projections increased. This demonstrates the impact of nigrostriatal lesions on dopaminergic-containing regions outside the nigrostriatal zone. These findings highlight the pro-locomotory effect, therapeutic potential, and plasticity of the A13.
+
+The role of the A13 region in locomotion in sham mice
+
We provide the first direct evidence of the photoactivation of the A13 being sufficient in driving locomotion. It is now evident that the pro-locomotor function of ZI extends further rostrally than previous work in caudal ZI (cZI) indicates (Mitrofanis, 2005): photoactivation of cZI neurons increased animal movement speed in prey capture (Zhao et al., 2019) and active avoidance (Hormigo et al., 2020). Previous data targeting the mZI region, including somatostatin (SOM+), calretinin (CR+), and vGlut2+ neurons, did not change locomotor distance travelled in the OFT (Li et al., 2021). In our work, there may be a combinatorial effect of multiple populations being photostimulated or targeting more medial populations in the ZI. Our results are consistent with in vivo calcium dynamics from CaMKIIα+ rostral ZI cells, which overlap the A13 showing subpopulations whose activity correlates with either movement speed or anxiety-related locations (Li et al., 2021). Enhancement of the A13 activity appears to modulate locomotor activity in naive mice differently from more lateral GABAergic ZI populations (dorsal and ventral ZI). Microinjection of GABAA receptor agonists, muscimol (Wardas et al., 1988) or etomidate (Chen et al., 2023), into the ZI either evokes severe catalepsy or a significant reduction in locomotor distance and velocity, respectively. Suppression of GABAergic ZI activity can either increase locomotion by microinjection of GABAA receptor antagonist bicuculline (Périer et al., 2002) or induce bradykinesia and akinesia by chemogenetic or optogenetic inhibitions in healthy naive mice (Chen et al., 2023).
+
Furthermore, our previous work showing A13 projections to the MLR is consistent with our observed photoactivation effects. We, and others, demonstrated that the A13 contains DA neurons, which may contribute to the observed effects, possibly via D1/5 receptor activation (Ryczko et al., 2016). A13 region photoactivation produces increased locomotor speed averaging 13 cm/s and improves descent times on the pole test. The enhanced ability to perform the pole test, which requires the animal to grasp the vertical pole to descend safely without falling, provides further evidence for the role of A13 region neurons in movements. Since A13 stimulation did not alter coordination during the task, it suggests a complex behavioral role consistent with its upstream location from the brainstem and extensive connectome. A13 photoactivation increases animal speed, duration, and distance travelled. Interestingly, the latency in A13 to observe increases in ongoing animal speed or to initiate locomotion is long in both 6-OHDA and sham mice (∼ 15 and 5 seconds, respectively). Second-long delays are often typical of sites upstream of the cuneiform, such as the dlPAG, which has a delay of several seconds (Tsang et al., 2021). The delays in locomotor initiation and context-specific integration following stimulation of upstream CnF targets may offer a therapeutic advantage for overcoming gait dysfunction.
+
+
+Photoactivation of the A13 reduces bradykinesia and akinesia in mouse PD models
+
Several studies have focused on the basal ganglia by targeting the subthalamic nucleus (Gradinaru et al., 2009; Yoon et al., 2014), endopeduncular nucleus (Moon et al., 2018; Yoon et al., 2020), SNc (Kravitz et al., 2010), striatum (Bordia et al., 2016; Ryan et al., 2018), cZI (Li et al., 2022), and motor cortex (Magno et al., 2019; Sanders and Jaeger, 2016; Valverde et al., 2020), or their projections in mouse models of PD. MLR subpopulations have been explored as a target for PD DBS with mixed results (Fougère et al., 2021; Masini and Kiehn, 2022). Recently, Li et al. found that cZI glutamatergic neurons were overactive after administering 6-OHDA into the striatum, and photoinhibition rescued the motor deficits (Li et al., 2022). Motor deficits in a 6-OHDA-induced PD mouse model were also ameliorated by chemogenetic and optogenetic activation of dorsal and ventral ZI GABAergic neurons (Chen et al., 2023). Here, we introduce a novel subpopulation in ZI (A13 cells) whose photoactivation rescued bradykinetic and akinetic deficits observed in the 6-OHDA mice.
+
Our work shows that A13 projections are affected at cortical and striatal levels following 6-OHDA, consistent with our observed changes in locomotor function. Over 28 days, there was a remarkable change in the afferent and efferent A13 regional connectome, despite the preservation of TH+ ZI cells. This is consistent with previous reports of widespread connectivity of the ZI (Mitrofanis, 2005). The preservation of A13 is expected since A13 lacks DAT expression (Bolton et al., 2015; Negishi et al., 2020; Sharma et al., 2018) and is spared from DAT-mediated toxicity of 6-OHDA (Dauer and Przedborski, 2003; Konnova et al., 2018; Simola et al., 2007). While A13 cells are spared following nigrostriatal degeneration, our work demonstrates its connectome is rewired. The ipsilateral afferent projections were markedly downregulated, while contralesional projecting afferents showed upregulation. In contrast, efferent projections showed less downregulation in the cortical subplate regions and bilateral upregulation of thalamic and hypothalamic efferents. Similar timeframes for anatomical and functional plasticity affecting neurons and astrocytes following an SNc or MFB 6-OHDA have been previously reported (Bosson et al., 2015; Perović et al., 2005; Requejo et al., 2020). Human PD brains that show degeneration of the SNc have a preserved A13 region, suggesting that our model, from this perspective, is externally valid (Matzuk and Saper, 1985).
+
Combined with photoactivation of the A13 region, we provide evidence for plasticity following damage to SNc. A previous brain-wide quantification of TH levels in the MPTP mouse model identified additional complexity in regulating central TH expression compared to conventional histological studies (Roostalu et al., 2019). Roostalu et al. reported decreased SNc TH+ cell numbers without a significant change in TH+ intensity in SNc and increased TH+ intensity in limbic regions such as the amygdala and hypothalamus (Roostalu et al., 2019). Likewise, we found no significant change in A13 TH+ cell counts. Still, there was a downstream shift in the distribution pattern of A13 efferents following nigrostriatal degeneration with a pullback on outputs to cortical and striatal subregions. This suggests A13 efferents are more distributed across the neuraxis than in sham mice. The preserved A13 efferents could provide compensatory dopaminergic innervation with collateralization mediated contralesionally and, in some subregions, ipsilesionally to increase the availability of extracellular dopamine. Considering A13-MLR efferents (Sharma et al., 2018) that remain preserved, photoactivation of glutamatergic MLR neurons alleviates motor deficits in the 6-OHDA mouse model (Fougère et al., 2021; Masini and Kiehn, 2022), and photoactivation of A13 somata promotes locomotion in 6-OHDA mice - hypotheses that warrant further investigation.
+
Several A13 efferent targets could be responsible for rotational asymmetry. In a unilateral 6-OHDA model, ipsiversive circling behaviour is indicative of intact striatal function on the contralesional side (Carey, 1991; Schwarting et al., 1991; Ungerstedt, 1971; Zetterström et al., 1986). Instead, the predictive value of a treatment is determined by contraversive circling mediated by increased dopamine receptor sensitivity on the ipsilesional striatal terminals (Costall et al., 1976; Lane et al., 2006). Thus, our data suggest that photoactivation of ipsilesional A13 has an overall additive effect on ipsiversive circling and represents a gain of function on the intact side that contributes to the magnitude of overall motor asymmetry against the lesioned side. Since A13 cells are preserved in PD, future therapies could use bilateral stimulations optimized for each side to minimize the overall motor asymmetry while ameliorating bradykinesia and akinesia.
+
With the induction of a 6-OHDA lesion, there is a change in the A13 connectome, characterized by a reduction in bidirectional connectivity with ipsilesional cortical regions. In rodent models, the motor cortices, including the M1 and M2 regions, can shape rotational asymmetry (Gradinaru et al., 2009; Magno et al., 2019; Sanders and Jaeger, 2016; Valverde et al., 2020). Activation of M1 glutamatergic neurons increases the rotational bias (Valverde et al., 2020), while M2 neuronal stimulation promotes contraversive rotations (Magno et al., 2019). Our data suggest that A13 photoactivation may have resulted in the inhibition of glutamatergic neurons in the contralesional M1. An alternative possibility is the activation of the contralesional M2 glutamatergic neurons, which would be expected to induce increased ipsilesional rotations (Magno et al., 2019). The ZI could generate rotational bias by A13 modulation of cZI glutamatergic neurons via incerto-incertal fibres (Ossowska, 2019; Power and Mitrofanis, 1999), which promotes asymmetries by activating the SNr (Li et al., 2022). The incerto-incertal interconnectivity has not been well studied, but the ZI has a large degree of interconnectivity (Sharma et al., 2018; Tsang et al., 2021) along all axes and between hemispheres (Power and Mitrofanis, 1999). However, this may only contribute minimally given that unilateral photoactivation of the A13 cells in sham mice failed to produce ipsiversive turning behavior while unilateral photoactivation of cZI glutamatergic neurons in sham animals was sufficient in generating ipsiversive turning behavior (Li et al., 2022). Another possibility involves the A13 region projections to the MLR. With the unknown downstream effects of A13 photoactivation, there may be modulation of the PPN neurons responsible for this turning behavior (Masini and Kiehn, 2022). The thigmotaxic behaviors suggest some effects may be mediated through dlPAG and CnF (Tsang et al., 2021), and recent work suggests the CnF as a possible therapeutic target (Fougère et al., 2021; Noga and Whelan, 2022). Since PD is a heterogeneous disease, our data provide another therapeutic target providing context-dependent relief from symptoms. This is important since PD severity, symptoms, and progression are patient-specific.
+
+
+Towards a preclinical model
+
To facilitate future translational work applying DBS to this region, we targeted the A13 region using AAV8-CamKII-mCherry viruses. The CaMKIIα promoter virus is beneficial because it is biased towards excitatory cells (Haery et al., 2019), narrowing the diversity of transfected A13 region neurons and in our hands, the viral spread was contained within the A13 region. Optogenetic strategies have been used to activate retinal cells in humans, partially restoring visual function and providing optimism that AAV-based viral strategies can be adapted in other human brain regions (Sahel et al., 2021). A more likely possibility for stimulation of deep nuclei is that DREADD technology could be adapted, which would not require any implants, but this remains a longer-term possibility. In the short-term, our work suggests that the A13 is a possible target for DBS. Gait dysfunction in PD is particularly difficult to treat, and indeed when DBS of subthalamic nucleus is deployed, a mixture of unilateral and bilateral approaches have been used (Lizarraga et al., 2016), along with stimulation of multiple targets (Stefani et al., 2007). This represents the heterogeneity of PD and underlines the need for considering multiple targets. In this regard, the identification of non-canonical dopaminergic pathways for the direct control of locomotion is promising (Figure 6). Our work highlights the A13 as a possible target, likely used in context and in concert with the activation of other identified targets.
+
+
+
Comparing descending dopamine pathways for locomotor control.
+
Simplified connectivity map for the 3 dopamine pathways. The first pathway is the classical VTA/SNc projection to the striatum, and the SNr/GPi projects to the MLR. The VTA/SNc also directly projects to the MLR (Ryczko et al., 2016). The mZI/A13 region projects dopaminergic projections to the MLR (Sharma et al., 2018). Canonical pathways are in black, while non-canonical pathways are in red.
+
+
+
+
+Limitations
+
Currently, few PD animal models are available that adequately model the progression and the extent of SNc cellular degeneration while meeting the face validity of motor deficits (Dauer and Przedborski, 2003; Konnova et al., 2018). While the 6-OHDA models fail to capture the age-dependent chronic degeneration observed in PD, it provides additional advantages in providing robust motor deficits with acute degeneration and identifying compensatory changes compared to the unlesioned side. Moreover, it resembles the unilateral onset (Hughes et al., 1992) and persistent asymmetry (Lee et al., 1995) of motor dysfunction in PD. Another option could be the MPTP mouse model, which offers the ease of systemic administration and translational value to primate models; however, the motor deficits are variable and lack the asymmetry observed in human patients (Hughes et al., 1992; Jagmag et al., 2015; Lee et al., 1995; Meredith and Rademacher, 2011). Despite these limitations, the neurotoxin-based mouse models, such as MPTP and 6-OHDA, offer greater SNc cell loss than genetic-based models; in the case of the 6-OHDA model, it captures many aspects of motor dysfunctions in PD (Dauer and Przedborski, 2003; Jagmag et al., 2015; Konnova et al., 2018; Simola et al., 2007).
+
+
+Conclusions
+
Parkinson’s disease involves areas outside the classic nigrostriatal axis. Our work demonstrates that the A13 region drives locomotor activity and rescues bradykinetic and akinetic deficits caused by dysfunctional DAergic transmission in the basal ganglia. We show that A13 region-evoked locomotion has therapeutic potential for improving gait in advanced PD. Widespread remodelling of the A13 region connectome is critical to our understanding of the effects of dopamine loss in PD models. In summary, our findings support an exciting role for the A13 region in locomotion with demonstrated benefits in a mouse PD model and contribute to our understanding of heterogeneity in PD.
+
+
+
+MATERIALS AND METHODS
+
+Animals
+
All care and experimental procedures were approved by the University of Calgary Health Sciences Animal Care Committee (Protocol #AC19-0035). C57BL/6 male mice 49 - 56 days old (weight: M = 31.7 g, SEM = 2.0 g) were group-housed (≤ four per cage) on a 12-h light/dark cycle (07:00 lights on - 19:00 lights off) with ad libitum access to food and water, as well as cat’s milk (Whiskas, Mars Canada Inc., Bolton, ON, Canada). Mice were randomly assigned to the groups described.
+
+
+Surgical Procedures
+
We established a well-validated unilateral 6-OHDA mediated Parkinsonian mouse model (Thiele et al., 2012) (Figure 1, Movie S4). 30 minutes before stereotaxic microinjections, mice were intraperitoneally injected with desipramine hydrochloride (2.5 mg/ml, Sigma-Aldrich) and pargyline hydrochloride (0.5 mg/ml, Sigma-Aldrich) at 10 ml/kg (0.9% sterile saline, pH 7.4) to enhance selectivity and efficacy of 6-OHDA induced lesions (Thiele et al., 2012). All surgical procedures were performed using aseptic techniques, and mice were anesthetized using isoflurane (1 - 2%) delivered by 0.4 L/min of medical-grade oxygen (Vitalair 1072, 100% oxygen).
+
Mice were stabilized on a stereotaxic apparatus. Small craniotomies were made above the medial forebrain bundle (MFB) and the A13 nucleus within one randomly assigned hemisphere. Stereotaxic microinjections were performed using a glass capillary (Drummond Scientific, PA, USA; Puller Narishige, diameter 15 – 20 mm) and a Nanoject II apparatus (Drummond Scientific, PA, USA). 240 nL of 6-OHDA (3.6 µg, 15.0 mg/mL; Tocris, USA) was microinjected into the MFB (AP −1.2 mm from bregma; ML ±1.1 mm; DV −5.0 mm from the dura). Sham mice received a vehicle solution (240 nL of 0.2% ascorbic acid in 0.9% saline; Tocris, USA).
+
+Whole Brain Experiments
+
For tracing purposes, a 50:50 mix of AAV8-CamKII-mCherry (Neurophotonics, Laval University, Quebec City, Canada, Lot #820, titre 2×1013 GC/ml) and AAVrg-CAG-GFP (Addgene, Watertown, MA, Catalogue #37825, Lot #V9234, titre ≥ 7×10¹² vg/mL) was injected ipsilateral to 6-OHDA injections at the A13 nucleus in all mice (AP −1.22 mm from bregma; ML ±0.4 mm; DV −4.5 mm from the dura, the total volume of 110 nL at a rate of 23 nl/sec). Post-surgery care was the same for both sham and 6-OHDA injected mice. The animals were sacrificed 29 days after surgery.
+
+
+Photoactivation Experiments
+
36.8 nL of AAVDJ-CaMKIIα-hChR2(H134R)-eYFP (UNC Stanford Viral Gene Core; Stanford, CA, US, Catalogue #AAV36; Lots #3081 and #6878, titres 1.9×1013 and 1.7×1013 GC/mL, respectively) or eYFP control virus (AAVDJ-CaMKIIα-eYFP; Lots #2958 & #5510, titres 7.64×1013 and 2.88×1013 GC/mL, respectively) was injected into the A13’s stereotaxic coordinates (Sharma et al., 2018). A mono-fibre cannula (Doric Lenses, Quebec, Canada, Catalogue #B280-2401-5, MFC_200/230-0.48_5mm_MF2.5_FLT) was implanted slowly 300 μm above the viral injection site. Metabond® Quick Adhesive Cement System (C&B, Parkell, Brentwood, NY, US) and Dentsply Repair Material (Dentsply International Inc., York, PA, USA) were used to fix the ferrule in place. Animals recovered from the viral surgery for 19 days before follow-up behavioral testing. Figure 1 illustrates the timeline of the behavioral tests.
+
+
+
+ChR2 photoactivation
+
A Laserglow Technologies 473 nm laser and driver (LRS-0473-GFM-00100-05, North York, ON, Canada) were used to generate the photoactivation for experiments. The laser was triggered with TTL pulses from either an A.M.P.I. Master-8 stimulator (Jerusalem, Israel) or an Open Ephys PulsePal (Sanworks, Rochester, NY, US) set to 20 Hz with 10-ms pulse width. All fibre optic implants were tested for laser power before implantation (Thorlabs, Saint-Laurent, QC, Canada; optical power sensor (S130C) and meter (PM100D)). The Stanford Optogenetics irradiance calculator was used to estimate the laser power for stimulation (“Stanford Optogenetics Resource Center,” n.d.). A 1×2 fibre-optic rotary joint (Doric Lenses, Quebec, Canada; FRJ_1x2i_FC-2FC_0.22) was used. The animals’ behaviors were recorded with an overhead camera (SuperCircuits, Austin, TX, US; FRJ_1x2i_FC-2FC_0.22; 720 x 480 resolution; 30 fps). The video was processed online (Cleversys, Reston, VA; TopScan V3.0) with a TTL signal output from a National Instruments 24-line digital I/O box (NI, Austin, TX, US; USB-6501) to the Master-8 stimulator.
+
+
+Behavioral Testing
+
+Open Field Test
+
Each mouse was placed in a square arena measuring 70 (W) x 70 (L) x 50 (H) cm with opaque walls and recorded for 30 minutes using a vertically mounted video camera (Model PC165DNR, Supercircuits, Austin, TX, USA; 30 fps). 19 days following surgery, mice were habituated to the open field test (OFT) arena with a patch cable attached for three days in 30 minute sessions to bring animals to a common baseline of activity. On experimental days, after animals were placed in the OFT, a one-minute-on-three-minutes-off paradigm was repeated five times following an initial ten minutes baseline activity. Locomotion was registered when mice travelled a minimum distance of 10 cm at 6 cm/s for 20 frames over a 30-frame segment. When the mouse velocity dropped below 6 cm/s for 20 frames, locomotion was recorded as ending. Bouts of locomotion relate to the number of episodes where the animal met these criteria. Velocity data were obtained from the frame-by-frame results and further processed in a custom Python script to detect instantaneous speeds greater than 2 cm/s (Masini and Kiehn, 2022). All animals that had validated targeting of the A13 region were included in the OFT data presented in the results section, except for one sham ChR2 animal, which showed grooming rather than the typical locomotor phenotypes.
+
+
+Pole Test
+
Mice were placed on a vertical wooden pole (50 cm tall and 1 cm diameter) facing upwards and then allowed to descend the pole into their home cage (Glajch et al., 2012). Animals were trained for three days and tested 2-5 days pre-surgery. Animals were acclimatized 21-22 days post-surgery under two conditions: without a patch cable and with the patch cable attached without photoactivation. On days 24-27, experimental trials were recorded with photoactivation. Video data were recorded for a minimum of three trials (Canon, Brampton, ON, Canada; Vixia HF R52; 1920 x 1080 resolution; 60 fps). A blinded scorer recorded the times for the following events: the hand release of the animal’s tail, the animal fully turning to descend the pole, and the animal reaching the base of the apparatus. Additionally, partial falls, where the animal slipped down the pole but did not reach the base, and full falls, where the animal fell to the base, were recorded separately. All validated animals were included in the quantified data, including the sham ChR2 animal that began grooming in the OFT upon photoactivation. This animal displayed proficiency in performing the PT during photoactivation. It started grooming upon completion of the task when photoactivation was on. One sham ChR2 animal was photostimulated at 1 mW since it would jump off the apparatus at higher stimulation intensities.
+
+
+
+Immunohistochemistry
+
+A13 and SNc region
+
Post hoc analysis of the tissue was performed to confirm the 6-OHDA lesion and validate the targeting of the A13 region. Following behavioral testing, animals underwent a photoactivation protocol to activate neurons below the fibre optic tip (Koblinger et al., 2018). Animals were placed in an OFT for ten minutes before receiving three minutes of photoactivation. Ten minutes later, the animals were returned to their home cage. 90 minutes post photoactivation, animals were deeply anaesthetised with isoflurane and then transcardially perfused with room temperature PBS followed by cold 4% paraformaldehyde (PFA) (Sigma-Aldrich, Catalogue #441244-1KG). The animals were decapitated, and the whole heads were incubated overnight in 4% PFA at 4°C before the fibre optic was removed and the brain removed from the skull. The brain tissue was post-fixed for another 6 - 12 hours in 4% PFA at 4°C then transferred to 30% sucrose solution for 48 - 72 hours. The tissue was embedded in VWR® Clear Frozen Section Compound (VWR International LLC, Radnor, PA, US) and sectioned coronally at 40 or 50 μm using a Leica cryostat set to −21°C (CM 1850 UV, Concord, ON, Canada). Sections from the A13 region (−0.2 to −2.0 mm past bregma) and the SNc (−2.2 to −4.0 past bregma) were collected and stored in PBS containing 0.02% (w/v) sodium azide (EM Science, Catalogue #SX0299-1, Cherry Hill, NJ, US) (Keith B. J. Franklin and Paxinos, 2008).
+
Immunohistochemistry staining was done on free-floating sections. The A13 sections were labelled for c-Fos, TH, and GFP (to enhance eYFP viral signal), and received a DAPI stain to identify nuclei. The SNc sections were stained with TH and DAPI (Table 1). Sections were washed in PBS (3 x 10 mins) then incubated in a blocking solution comprised of PBS containing 0.5% Triton X-100 (Sigma-Aldrich, Catalogue #X100-500ML, St. Louis, MO, US) and 5% donkey serum (EMD Millipore, Catalogue #S30-100ML, Billerica, MA, USA) for 1 hour. This was followed by overnight (for SNc sections) or 24-hour (for A13 sections) incubation in a 5% donkey serum PBS primary solution at room temperature. On day 2, the tissue was washed in PBS (3 x 10 mins) before being incubated in a PBS secondary solution containing 5% donkey serum for 2 hours (for SNc tissue) or 4 hours (for A13 tissue). The secondary was washed with a PBS solution containing 1:1000 DAPI for 10 mins, followed by a final set of PBS washes (3 x 10 mins). Tissue was mounted on Superfrost® micro slides (VWR, slides, Radnor, PA, US) with mounting media (Vectashield®, Vector Laboratories Inc., Burlingame, CA, US), covered with #1 coverslips (VWR, Radnor, PA, US) then sealed.
+
+
+
List of antibodies used for immunohistochemical staining of the A13 and SNc regions, as well as the whole brain.
+
+
+
+
+Whole Brain
+
Mice were deeply anesthetized with isoflurane and transcardially perfused with PBS, followed by 4% PFA. To prepare for whole brain imaging, brains were first extracted and postfixed overnight in 4% PFA at 4°C. The next day, a modified iDISCO method (Renier et al., 2014) was used to clear the samples and perform quadruple immunohistochemistry in whole brains. The modifications include prolonged incubation and the addition of SDS for optimal labelling. The antibodies used are listed in Table 1 and the protocol is provided in Table 2.
All tissue was initially scanned with an Olympus VS120-L100 Virtual Slide Microscope (UPlanSApo, 10x and 20x, NA = 0.4 and 0.75). Standard excitation and emission filter cube sets were used (DAPI, FITC, TRITC, Cy5), and images were acquired using an Orca Flash 4.0 sCMOS monochrome camera (Hamamatsu, Bridgewater Township, NJ, US). For c-Fos immunofluorescence, A13 sections of the tissue were imaged with a Leica SP8 FALCON (FAst Lifetime CONtrast) scanning confocal microscope equipped with a tunable laser and using a 63x objective (HC PlanApo, NA = 1.40).
+
SNc images were imported into Adobe Illustrator, where the SNc (Fougère et al., 2021), including the pars lateralis (SNl), was delineated using the TH immunostaining together with the medial lemniscus and cerebral peduncle as landmarks (bregma −3.09 and −3.68) (Iancu et al., 2005; Keith B. J. Franklin and Paxinos, 2008; Stott and Barker, 2014). Cell counts were obtained using a semi-automated approach using an Ilastik (v1.4.0b15) (Berg et al., 2019) trained model followed by corrections by a blinded counter (Fougère et al., 2021; Iancu et al., 2005). Targeting was confirmed on the 10x overview scans of the A13 region tissue by the presence of eYFP localized in the mZI around the A13 TH+ nucleus, the fibre optic tip being visible near the mZI and A13 nucleus, and the presence of c-Fos positive cells in ChR2+ tissue. C-Fos expression colocalization within the A13 region was performed using confocal images. The mZI & A13 region was identified with the 3rd ventricle and TH expression as markers (Keith B. J. Franklin and Paxinos, 2008).
+
+
+Whole Brain Experiments
+
Cleared whole brain samples were imaged using a light-sheet microscope (LaVision Biotech UltraMicroscope, LaVision, Bielefeld, Germany) with an Olympus MVPLAPO 2x objective with 4x optical zoom (NA = 0.475) and a 5.7 mm dipping cap that is adjusted for the high refractive index of 1.56. The brain samples were imaged in an ethyl cinnamate medium to match the refractive indices and illuminated by three sheets of light bilaterally. Each light sheet was 5 µm thick, and the width was set at 30% to ensure sufficient illumination at the centroid of the sample. Laser power intensities and chromatic aberration corrections used for each laser were as follows: 10% power for 488 nm laser, 5% power for 561 nm laser with 780 nm correction, 40% power for 640 nm laser with 960 nm correction, and 100% power for 785 nm laser with 1,620 nm correction. Each sample was imaged coronally in 8 by 6 squares with 20% overlap (10,202 µm by 5,492 µm in total) and a z-step size of 15 µm (xyz resolution = 0.813 µm x 0.813 µm x 15 µm). While an excellent choice for our work, confocal microscopy offers better resolution at the expense of time. To gain a better resolution using a light-sheet microscope in select regions (eg. SNc and A13 cells), we increased the optical zoom to 6.3x.
+
+
+A13 Connectome Analysis
+
Images were processed using ImageJ software (Schneider et al., 2012). Raw images were stitched, and a z-encoded maximum intensity projection across a 90 µm thick optical section was obtained across each brain. 90 µm sections were chosen because the 2008 Allen reference atlas images are spaced out at around 100 µm. Brains with insufficient quality in labeling were excluded from analysis (n = 1 of three sham and n = 3 of six 6-OHDA mice). Instructions for identifying YFP+ or TH+ cells to annotate were provided to the manual counters. YFP+ and TH+ cells were manually counted using the Cell Counter Plug-In (ImageJ). mCherry+ fibers were segmented semi-automatically using Ilastik software (Berg et al., 2019) and quantified using particle analysis in ImageJ. Images and segmentations were imported into WholeBrain software to be registered with the 2008 Allen reference atlas (Fürth et al., 2018). The TO-PRO™-3 and TH channels were used as reference channels to register each section to a corresponding atlas image. ImageJ quantifications of cell and fiber segmentations were exported in XML formats and registered using WholeBrain software. To minimize the influence of experimental variation on the total labeling of neurons and fibers, the afferent cell counts or efferent fiber areas in each brain region were column divided by the total number found in a brain to obtain the proportion of total inputs and outputs. Connectome analyses were performed using custom R scripts (L. H. Kim et al., 2021). For interregional correlation analyses, the data were normalized to a log10 value to reduce variability and bring brain regions with high and low proportions of cells and fibers to a similar scale. The consistency of afferent and efferent proportions between mice was compared in a pairwise manner using Spearman’s correlation (Figure S5).
+
+
+Quantification of 6-OHDA mediated TH<sup>+</sup> cell loss
+
The percentage of TH+ cell loss was quantified to confirm 6-OHDA mediated SNc lesions. TH+ cells within ZI, VTA and SNc areas from 90 µm thick optical brain slice images (AP: −0.655 to −3.88 mm from bregma) were manually counted by two blinded counters (n = 3 sham and n = 6 6-OHDA mice; ZI region in 2 of 6 6-OHDA mice were excluded due to presence of abnormal scarring/healing at the injection site of viruses). Subsequently, WholeBrain software (Fürth et al., 2018) was used to register and tabulate TH+ cells in the contralesional and ipsilesional brain regions of interest. Counts obtained from the two counters were averaged per region. The percentage of TH+ cell loss was calculated by dividing the difference in counts between contralesional and ipsilesional sides by the contralesional side count and multiplying by 100%.
+
+
+
+Statistical analyses
+
All data were tested for normality using a Shapiro-Wilk test to determine the most appropriate statistical tests. The percent ipsilesional TH+ neuron loss within the SNc as defined above using a Pearson correlation (Fougère et al., 2021) was used to ascertain the effect of the 6-OHDA lesion on behavior. A Wilcoxon rank-sum test was performed for comparisons within subjects at two timepoints where normality failed, and the central limit theorem could not be applied. The two groups were compared using an unpaired t-test with Welch’s correction. A mixed model ANOVA (MM ANOVA) was used to compare the effects of group type, injection type and time. Additionally, Mauchly’s test of sphericity was performed to account for differences in variability within the repeated measures design. A Greenhouse Geisser correction was applied to all ANOVAs where Mauchly’s test was significant for RM and MM ANOVAs. The post hoc multiple comparisons were run when the respective ANOVAs reached significance using Dunnett’s or Dunn’s tests for repeated measures of parametric and non-parametric tests, respectively. The pre-stimulation timepoints were used as the control time point to determine if stimulation altered behavior. A Bonferroni correction was added for post hoc comparisons following a MM ANOVA between groups at given time points to control for alpha value inflation. All correlations, t-tests, and ANOVAs were performed, and graphs were created using Prism version 9.3.1 (Graphpad) or SPSS (IBM, 28.0.1.0). Full statistical reporting is in Supplemental Statistics.xls.
+
+
+Figures
+
Figures were constructed using Adobe Photoshop, Illustrator, and Biorender.
+
+
+
+
+
+Author Contributions
+
LHK and AL performed experiments and prepared figures. MAT and PJW edited figures. LHK, AL, ZHT and PJW conceived and designed the research and interpreted the results. PJW procured funding for the experiments. SS and AL performed surgeries for lesions, optogenetic experiments and conducted behavioral experiments. LHK and SEAE optimized light-sheet imaging. MAT, ST, and CM performed manual cell counting. TC performed analysis and prepared figures on gait analysis. LHK, AL, TC, ZHT, and PJW drafted the manuscript. All authors reviewed and approved the final version of the manuscript.
+
+
+Competing Interest Statement
+
None.
+
+
+ACKNOWLEDGEMENTS
+
We would like to acknowledge support from Whelan and Kiss Labs and technical support from Hotchkiss Brain Institute Advanced Microscopy Platform Core Facility, Cumming School of Medicine Optogenetics Platform Core Facility and Drs. David Elliot, Jonathan Epp, Young Ou, and Lothar Resch. We acknowledge studentships from Parkinson Alberta (LHK), Parkinson Canada (LHK), Canadian Open Neuroscience Platform (AL), Cumming School of Medicine (AL, LHK), Faculty of Graduate Studies (AL, LHK), and the Faculty of Veterinary Medicine (CM, ST). This research was supported by grants to PJW provided by a Canadian Institutes of Health Research Project Grant (PJT-173511), Wings for Life, NSERC (RGPIN/04394-2019) as well as ZHTK from NSERC (RPGIN/04126-2017).
+
+
+DATA AVAILABILITY
+
All datasets and code will be made available on a public repository.
+
+
+REFERENCES
+BergS, KutraD, KroegerT, StraehleCN, KauslerBX, HauboldC, SchieggM, AlesJ, BeierT, RudyM, ErenK, CervantesJI, XuB, BeuttenmuellerF, WolnyA, ZhangC, KoetheU, HamprechtFA, KreshukA. 2019. ilastik: interactive machine learning for (bio)image analysis. Nat Methods16:1226–1232.
+BlomstedtP, Stenmark PerssonR, HarizG-M, LinderJ, FredricksA, HäggströmB, PhilipssonJ, ForsgrenL, HarizM. 2018. Deep brain stimulation in the caudal zona incerta versus best medical treatment in patients with Parkinson’s disease: a randomised blinded evaluation. J Neurol Neurosurg Psychiatry89:710–716.
+BoixJ, PadelT, PaulG. 2015. A partial lesion model of Parkinson’s disease in mice--characterization of a 6-OHDA-induced medial forebrain bundle lesion. Behav Brain Res284:196–206.
+BoltonAD, MurataY, KirchnerR, KimS-Y, YoungA, DangT, YanagawaY, Constantine-PatonM. 2015. A Diencephalic Dopamine Source Provides Input to the Superior Colliculus, where D1 and D2 Receptors Segregate to Distinct Functional Zones. Cell Rep13:1003–1015.
+BordiaT, PerezXA, HeissJ, ZhangD, QuikM. 2016. Optogenetic activation of striatal cholinergic interneurons regulates L-dopa-induced dyskinesias. Neurobiol Dis91:47–58.
+BossonA, BoisseauS, BuissonA, SavastaM, AlbrieuxM. 2015. Disruption of dopaminergic transmission remodels tripartite synapse morphology and astrocytic calcium activity within substantia nigra pars reticulata. Glia63:673–683.
+BraakH, Del TrediciK, RübU, de VosRAI, Jansen SteurENH, BraakE.2003. Staging of brain pathology related to sporadic Parkinson’s disease. Neurobiol Aging24:197–211.
+CaireF, RanouxD, GuehlD, BurbaudP, CunyE. 2013. A systematic review of studies on anatomical position of electrode contacts used for chronic subthalamic stimulation in Parkinson’s disease. Acta Neurochir155:1647–54; discussion 1654.
+CareyRJ. 1991. Chronic L-dopa treatment in the unilateral 6-OHDA rat: evidence for behavioral sensitization and biochemical tolerance. Brain Res568:205–214.
+CenciMA, BjörklundA. 2020. Animal models for preclinical Parkinson’s research: An update and critical appraisal. Prog Brain Res252:27–59.
+ChenF, QianJ, CaoZ, LiA, CuiJ, ShiL, XieJ.2023. Chemogenetic and optogenetic stimulation of zona incerta GABAergic neurons ameliorates motor impairment in Parkinson’s disease. iScience26:107149.
+ChoiEA, McNallyGP. 2017. Paraventricular Thalamus Balances Danger and Reward. J Neurosci37:3018–3029.
+CohenJ.1988. Statistical power analysis for the behavioral sciencesNew York. NY: Academic54.
+CostallB, NaylorRJ, PycockC. 1976. Non-specific supersensitivity of striatal dopamine receptors after 6-hydroxydopamine lesion of the nigrostriatal pathway. Eur J Pharmacol35:276–283.
+DauerW, PrzedborskiS. 2003. Parkinson’s disease: mechanisms and models. Neuron39:889–909.
+EatonMJ, WagnerCK, MooreKE, LookinglandKJ. 1994. Neurochemical identification of A13 dopaminergic neuronal projections from the medial zona incerta to the horizontal limb of the diagonal band of Broca and the central nucleus of the amygdala. Brain Res659:201–207.
+FerrayeMU, DebûB, FraixV, GoetzL, ArdouinC, YelnikJ, Henry-LagrangeC, SeigneuretE, PiallatB, KrackP, Le BasJ-F, BenabidA-L, ChabardèsS, PollakP.2010. Effects of pedunculopontine nucleus area stimulation on gait disorders in Parkinson’s disease. Brain133:205–214.
+FougèreM, van der ZouwenCI, BoutinJ, NeszvecskoK, SarretP, RyczkoD.2021. Optogenetic stimulation of glutamatergic neurons in the cuneiform nucleus controls locomotion in a mouse model of Parkinson’s disease. Proc Natl Acad Sci U S A118. doi:10.1073/pnas.2110934118
+FürthD, VaissièreT, TzortziO, XuanY, MärtinA, LazaridisI, SpigolonG, FisoneG, TomerR, DeisserothK, CarlénM, MillerCA, RumbaughG, MeletisK. 2018. An interactive framework for whole-brain maps at cellular resolution. Nat Neurosci21:139–149.
+GlajchKE, FlemingSM, SurmeierDJ, OstenP. 2012. Sensorimotor assessment of the unilateral 6-hydroxydopamine mouse model of Parkinson’s disease. Behav Brain Res230:309–316.
+GlicksteinM, SteinJ. 1991. Paradoxical movement in Parkinson’s disease. Trends Neurosci14:480–482.
+GoldowitzD. 2010. Allen Reference Atlas. A Digital Color Brain Atlas of the C57BL/6J Male Mouse - by H. W. Dong. Genes, Brain and Behavior. doi:10.1111/j.1601-183x.2009.00552.x
+GradinaruV, MogriM, ThompsonKR, HendersonJM, DeisserothK. 2009. Optical deconstruction of parkinsonian neural circuitry. Science324:354–359.
+GutNK, WinnP. 2015. Deep Brain Stimulation of Different Pedunculopontine Targets in a Novel Rodent Model of Parkinsonism. Journal of Neuroscience. doi:10.1523/jneurosci.3646-14.2015
+HaeryL, DevermanBE, MathoKS, CetinA, WoodardK, CepkoC, GuerinKI, RegoMA, ErsingI, BachleSM, KamensJ, FanM. 2019. Adeno-Associated Virus Technologies and Methods for Targeted Neuronal Manipulation. Front Neuroanat13:93.
+HamaniC, MoroE, LozanoAM. 2011. The pedunculopontine nucleus as a target for deep brain stimulation. J Neural Transm118:1461–1468.
+HansenHH, RoostaluU, Hecksher-SørensenJ. 2020. Whole-brain three-dimensional imaging for quantification of drug targets and treatment effects in mouse models of neurodegenerative diseases. Neural Regeneration Res15:2255–2257.
+HoffmanBJ, PalkovitsM, PaiakK, HamsonSR, MezeyÉ. 1997. Regulation of Dopamine Transporter mRNA Levels in the Central Nervous System In: Goldstein DS, Eisenhofer G, McCarty R, editors. Advances in Pharmacology. Academic Press. pp. 202–206.
+HormigoS, ZhouJ, Castro-AlamancosMA. 2020. Zona Incerta GABAergic Output Controls a Signaled Locomotor Action in the Midbrain Tegmentum. eNeuro7. doi:10.1523/ENEURO.0390-19.2020
+HughesAJ, Ben-ShlomoY, DanielSE, LeesAJ. 1992. What features improve the accuracy of clinical diagnosis in Parkinson’s disease: a clinicopathologic study. Neurology42:1142–1146.
+IancuR, MohapelP, BrundinP, PaulG. 2005. Behavioral characterization of a unilateral 6-OHDA-lesion model of Parkinson’s disease in mice. Behav Brain Res162:1–10.
+JagmagSA, TripathiN, ShuklaSD, MaitiS, KhuranaS. 2015. Evaluation of Models of Parkinson’s Disease. Front Neurosci9:503.
+KeithB. J.FranklinMA, PaxinosG. 2008. The Mouse Brain in Stereotaxic Coordinates, Compact: The Coronal Plates and Diagrams. Elsevier Science.
+KimbroughA, LurieDJ, CollazoA, KreifeldtM, SidhuH, MacedoGC, D’EspositoM, ContetC, GeorgeO. 2020. Brain-wide functional architecture remodeling by alcohol dependence and abstinence. Proc Natl Acad Sci U S A117:2149–2159.
+KimL, ChomiakT, TranMA, TamS, McPhersonC, EatonSEA, OuY, KissZHT, WhelanPJ.2021. Global remodelling of afferent and efferent projections of the A13 region following unilateral nigrostriatal degeneration using 6-hydroxydopamineNeuroscience Meeting Planner. Presented at the Society for Neuroscience.
+KimLH, ChomiakT, TranM, TamS, McPhersonC, EatonSEA, OuY, ReschL, KissZHT, WhelanPJ.2021. Substantia nigra degradation results in widespread changes in medial zona incerta afferent and efferent connectomics. bioRxiv.
+KimLH, SharmaS, SharplesSA, MayrKA, KwokCHT, WhelanPJ. 2017. Integration of Descending Command Systems for the Generation of Context-Specific Locomotor Behaviors. Front Neurosci11:581.
+KishSJ, TongJ, HornykiewiczO, RajputA, ChangL-J, GuttmanM, FurukawaY. 2008. Preferential loss of serotonin markers in caudate versus putamen in Parkinson’s disease. Brain131:120–131.
+KoblingerK, Jean-XavierC, SharmaS, FüzesiT, YoungL, EatonSEA, KwokCHT, BainsJS, WhelanPJ. 2018. Optogenetic Activation of A11 Region Increases Motor Activity. Front Neural Circuits12:86.
+KonnovaEA, Translational NeurogeneticsUnit, Wallenberg NeuroscienceCenter, Lund University, Lund, Sweden, Swanberg M, Translational NeurogeneticsUnit, Wallenberg NeuroscienceCenter, Lund University, Lund, Sweden. 2018. Animal models of Parkinson’s diseaseParkinson’s Disease: Pathogenesis and Clinical Aspects. Codon Publications. pp. 83–106.
+KravitzAV, FreezeBS, ParkerPRL, KayK, ThwinMT, DeisserothK, KreitzerAC. 2010. Regulation of parkinsonian motor behaviours by optogenetic control of basal ganglia circuitry. Nature466:622–626.
+LaneEL, CheethamSC, JennerP. 2006. Does contraversive circling in the 6-OHDA-lesioned rat indicate an ability to induce motor complications as well as therapeutic effects in Parkinson’s disease?Exp Neurol197:284–290.
+LeeCS, SchulzerM, MakE, HammerstadJP, CalneS, CalneDB. 1995. Patterns of asymmetry do not change over the course of idiopathic parkinsonism: implications for pathogenesis. Neurology45:435–439.
+LiL-X, LiY-L, WuJ-T, SongJ-Z, LiX-M. 2022. Glutamatergic Neurons in the Caudal Zona Incerta Regulate Parkinsonian Motor Symptoms in Mice. Neurosci Bull38:1–15.
+LimS-Y, FoxSH, LangAE. 2009. Overview of the extranigral aspects of Parkinson disease. Arch Neurol66:167–172.
+LiS, ShiY, KirouacGJ. 2014. The hypothalamus and periaqueductal gray are the sources of dopamine fibers in the paraventricular nucleus of the thalamus in the rat. Front. Neuroanat8:136. 10.3389/fnana.2014.00136
+LizarragaKJ, JagidJR, LucaCC. 2016. Comparative effects of unilateral and bilateral subthalamic nucleus deep brain stimulation on gait kinematics in Parkinson’s disease: a randomized, blinded study. J Neurol263:1652–1656.
+LiZ, RizziG, TanKR. 2021. Zona incerta subpopulations differentially encode and modulate anxiety. Sci Adv7:eabf6709.
+MagnoLAV, Tenza-FerrerH, CollodettiM, AguiarMFG, Rodrigues APC, da Silva RS, Silva J do P, Nicolau NF, Rosa DVF, Birbrair A, Miranda DM, Romano-Silva MA. 2019. Optogenetic Stimulation of the M2 Cortex Reverts Motor Dysfunction in a Mouse Model of Parkinson’s Disease. J Neurosci39:3234–3248.
+MasiniD, KiehnO. 2022. Targeted activation of midbrain neurons restores locomotor function in mouse models of parkinsonism. Nat Commun13:504.
+MatsuuraK, KabutoH, MakinoH, OgawaN. 1997. Pole test is a useful method for evaluating the mouse movement disorder caused by striatal dopamine depletion. J Neurosci Methods73:45–48.
+MatzukMM, SaperCB. 1985. Preservation of hypothalamic dopaminergic neurons in Parkinson’s disease. Ann Neurol18:552–555.
+MeredithGE, RademacherDJ. 2011. MPTP mouse models of Parkinson’s disease: an update. J Parkinsons Dis1:19–33.
+MessanviF, Eggens-MeijerE, RoozendaalB, van der WantJJ.2013. A discrete dopaminergic projection from the incertohypothalamic A13 cell group to the dorsolateral periaqueductal gray in rat. Front Neuroanat7:41.
+MitrofanisJ. 2005. Some certainty for the “zone of uncertainty”? Exploring the function of the zona incerta. Neuroscience130:1–15.
+MitrofanisJ, MikuleticL. 1999. Organisation of the cortical projection to the zona incerta of the thalamus. J Comp Neurol412:173–185.
+MokD, MogensonGJ. 1986. Contribution of zona incerta to osmotically induced drinking in rats. Am J Physiol251:R823–32.
+MoonHC, WonSY, KimEG, KimHK, ChoCB, ParkYS. 2018. Effect of optogenetic modulation on entopeduncular input affects thalamic discharge and behavior in an AAV2-α-synuclein-induced hemiparkinson rat model. Neurosci Lett662:129–135.
+MoriyaS, YamashitaA, MasukawaD, SetoyamaH, HwangY, YamanakaA, KuwakiT. 2020. Involvement of A13 dopaminergic neurons located in the zona incerta in nociceptive processing: a fiber photometry study. Mol Brain13:60.
+MoroE, HamaniC, PoonY-Y, Al-KhairallahT, DostrovskyJO, HutchisonWD, LozanoAM. 2010. Unilateral pedunculopontine stimulation improves falls in Parkinson’s disease. Brain133:215–224.
+NegishiK, PayantMA, SchumackerKS, WittmannG, ButlerRM, LechanRM, SteinbuschHWM, KhanAM, CheeMJ. 2020. Distributions of hypothalamic neuron populations coexpressing tyrosine hydroxylase and the vesicular GABA transporter in the mouse. J Comp Neurol528:1833–1855.
+NogaBR, WhelanPJ. 2022. The Mesencephalic Locomotor Region: Beyond Locomotor Control. Front Neural Circuits16:884785.
+NonnekesJ, BereauM, BloemBR. 2020. Freezing of Gait and Its Levodopa Paradox. JAMA Neurol77:287–288.
+NonnekesJ, SnijdersAH, NuttJG, DeuschlG, GiladiN, BloemBR. 2015. Freezing of gait: a practical approach to management. Lancet Neurol14:768–778.
+OgawaN, HiroseY, OharaS, OnoT, WatanabeY. 1985. A simple quantitative bradykinesia test in MPTP-treated mice. Res Commun Chem Pathol Pharmacol50:435–441.
+OgundeleOM, LeeCC, FrancisJ. 2017. Thalamic dopaminergic neurons projects to the paraventricular nucleus-rostral ventrolateral medulla/C1 neural circuit. The Anatomical Record. doi:10.1002/ar.23528
+OkunMS, FooteKD. 2010. Parkinson’s disease DBS: what, when, who and why? The time has come to tailor DBS targets. Expert Rev Neurother10:1847–1857.
+OssowskaK. 2019. Zona incerta as a therapeutic target in Parkinson’s disease. Journal of Neurology. doi:10.1007/s00415-019-09486-8
+PeoplesC, SpanaS, AshkanK, BenabidA-L, StoneJ, BakerGE, MitrofanisJ. 2012. Photobiomodulation enhances nigral dopaminergic cell survival in a chronic MPTP mouse model of Parkinson’s disease. Parkinsonism Relat Disord18:469–476.
+Perez-LloretS, BarrantesFJ. 2016. Deficits in cholinergic neurotransmission and their clinical correlates in Parkinson’s disease. NPJ Parkinsons Dis2:16001.
+PérierC, TremblayL, FégerJ, HirschEC. 2002. Behavioral consequences of bicuculline injection in the subthalamic nucleus and the zona incerta in rat. J Neurosci22:8711–8719.
+PérierC, VilaM, FégerJ, AgidY, HirschEC. 2000. Functional activity of zona incerta neurons is altered after nigrostriatal denervation in hemiparkinsonian rats. Exp Neurol162:215–224.
+PerovićM, MladenovićA, RakićL, RuzdijićS, KanazirS. 2005. Increase of GAP-43 in the rat cerebellum following unilateral striatal 6-OHDA lesion. Synapse56:170–174.
+PlahaP, KhanS, GillSS. 2008. Bilateral stimulation of the caudal zona incerta nucleus for tremor control. J Neurol Neurosurg Psychiatry79:504–513.
+PowerBD, MitrofanisJ. 1999. Evidence for extensive inter-connections within the zona incerta in rats. Neurosci Lett267:9–12.
+RenierN, WuZ, SimonDJ, YangJ, ArielP, Tessier-LavigneM. 2014. iDISCO: a simple, rapid method to immunolabel large tissue samples for volume imaging. Cell159:896–910.
+RequejoC, López-de-IpiñaK, Ruiz-OrtegaJÁ, FernándezE, CalvoPM, Morera-HerrerasT, MiguelezC, Cardona-GrifollL, CepedaH, UgedoL, LafuenteJV. 2020. Changes in Day/Night Activity in the 6-OHDA-Induced Experimental Model of Parkinson’s Disease: Exploring Prodromal Biomarkers. Front Neurosci14:590029.
+RoostaluU, SalinasCBG, ThorbekDD, SkytteJL, FabriciusK, BarkholtP, JohnLM, JurtzVI, KnudsenLB, JelsingJ, VrangN, HansenHH, Hecksher-SørensenJ. 2019. Quantitative whole-brain 3D imaging of tyrosine hydroxylase-labeled neuron architecture in the mouse MPTP model of Parkinson’s disease. Dis Model Mech12. doi:10.1242/dmm.042200
+RyanMB, Bair-MarshallC, NelsonAB. 2018. Aberrant Striatal Activity in Parkinsonism and Levodopa-Induced Dyskinesia. Cell Rep23:3438–3446.e5.
+RyczkoD, ConeJJ, AlpertMH, GoetzL, AuclairF, DubéC, ParentM, RoitmanMF, AlfordS, DubucR. 2016. A descending dopamine pathway conserved from basal vertebrates to mammals. Proc Natl Acad Sci U S A113:E2440–9.
+SahelJ-A, Boulanger-ScemamaE, PagotC, ArleoA, GalluppiF, MartelJN, DelauxA, de Saint AubertJ-B, de MontleauC, GutmanE, AudoI, DuebelJ, PicaudS, DalkaraD, BlouinL, TaielM, RoskaB.2021. Partial recovery of visual function in a blind patient after optogenetic therapy. Nat Med27:1223–1229.
+SandersTH, JaegerD. 2016. Optogenetic stimulation of cortico-subthalamic projections is sufficient to ameliorate bradykinesia in 6-ohda lesioned mice. Neurobiol Dis95:225–237.
+SangheraMK, Anselmo-FranciJ, McCannSM. 1991. Effect of Medial Zona Incerta Lesions on the Ovulatory Surge of Gonadotrophins and Prolactin in the Rat. Neuroendocrinology. doi:10.1159/000125931
+SangheraMK, GradyS, SmithW, WoodwardDJ, PorterJC. 1991. Incertohypothalamic A13 dopamine neurons: effect of gonadal steroids on tyrosine hydroxylase. Neuroendocrinology53:268–275.
+ScattonB, Javoy-AgidF, RouquierL, DuboisB, AgidY. 1983. Reduction of cortical dopamine, noradrenaline, serotonin and their metabolites in Parkinson’s disease. Brain Res275:321–328.
+SchneiderCA, RasbandWS, EliceiriKW. 2012. NIH Image to ImageJ: 25 years of image analysis. Nat Methods9:671–675.
+SchwartingRK, BonatzAE, CareyRJ, HustonJP. 1991. Relationships between indices of behavioral asymmetries and neurochemical changes following mesencephalic 6-hydroxydopamine injections. Brain Res554:46–55.
+SharmaS, KimLH, MayrKA, ElliottDA, WhelanPJ. 2018. Parallel descending dopaminergic connectivity of A13 cells to the brainstem locomotor centers. Sci Rep8:7972.
+ShawVE, SpanaS, AshkanK, BenabidA-L, StoneJ, BakerGE, MitrofanisJ. 2010. Neuroprotection of midbrain dopaminergic cells in MPTP-treated mice after near-infrared light treatment. J Comp Neurol518:25–40.
+SimolaN, MorelliM, CartaAR. 2007. The 6-hydroxydopamine model of Parkinson’s disease. Neurotox Res11:151–167.
+SitaLV, EliasCF, BittencourtJC. 2007. Connectivity pattern suggests that incerto-hypothalamic area belongs to the medial hypothalamic system. Neuroscience148:949–969.
+Stanford Optogenetics Resource Center. n.d. https://web.stanford.edu/group/dlab/cgi-bin/graph/chart.php
+StefaniA, LozanoAM, PeppeA, StanzioneP, GalatiS, TropepiD, PierantozziM, BrusaL, ScarnatiE, MazzoneP. 2007. Bilateral deep brain stimulation of the pedunculopontine and subthalamic nuclei in severe Parkinson’s disease. Brain130:1596–1607.
+StottSRW, BarkerRA. 2014. Time course of dopamine neuron loss and glial response in the 6-OHDA striatal mouse model of Parkinson’s disease. Eur J Neurosci39:1042–1056.
+ThevathasanW, DebuB, AzizT, BloemBR, BlahakC, ButsonC, CzerneckiV, FoltynieT, FraixV, GrabliD, JointC, LozanoAM, OkunMS, OstremJ, PaveseN, SchraderC, TaiC-H, KraussJK, MoroE, Movement Disorders Society PPN DBS Working Groupin collaboration with the World Society for Stereotactic and Functional Neurosurgery. 2018. Pedunculopontine nucleus deep brain stimulation in Parkinson’s disease: A clinical review. Mov Disord33:10–20.
+ThieleSL, WarreR, NashJE. 2012. Development of a unilaterally-lesioned 6-OHDA mouse model of Parkinson’s disease. J Vis Exp. doi:10.3791/3234
+TsangE, OrlandiniC, SurekaR, CrevennaAH, PerlasE, PrankerdI, MasferrerME, GrossCT.2021. Induction of flight via midbrain projections to the cuneiform nucleus. bioRxiv. doi:10.1101/2021.12.21.473683
+UngerstedtU. 1971. Striatal dopamine release after amphetamine or nerve degeneration revealed by rotational behaviour. Acta Physiol Scand Suppl367:49–68.
+ValverdeS, VandecasteeleM, PietteC, DerousseauxW, GangarossaG, Aristieta ArbelaizA, TouboulJ, DegosB, VenanceL. 2020. Deep brain stimulation-guided optogenetic rescue of parkinsonian symptoms. Nat Commun11:2388.
+VenkataramanA, HunterSC, DhinojwalaM, GhebrezadikD, GuoJ, InoueK, YoungLJ, DiasBG. 2021. Incerto-thalamic modulation of fear via GABA and dopamine. Neuropsychopharmacology46:1658–1668.
+WangX, ChouX-L, ZhangLI, TaoHW. 2020. Zona Incerta: An Integrative Node for Global Behavioral Modulation. Trends Neurosci43:82–87.
+WardasJ, OssowskaK, WolfarthS. 1988. Evidence for the independent role of GABA synapses of the zona incerta-lateral hypothalamic region in haloperidol-induced catalepsy. Brain Res462:378–382.
+YoonHH, NamM-H, ChoiI, MinJ, JeonSR. 2020. Optogenetic inactivation of the entopeduncular nucleus improves forelimb akinesia in a Parkinson’s disease model. Behav Brain Res386:112551.
+YoonHH, ParkJH, KimYH, MinJ, HwangE, LeeCJ, SuhJ-KF, HwangO, JeonSR. 2014. Optogenetic inactivation of the subthalamic nucleus improves forelimb akinesia in a rat model of Parkinson disease. Neurosurgery74:533–40; discussion 540–1.
+ZetterströmT, Herrera-MarschitzM, UngerstedtU. 1986. Simultaneous measurement of dopamine release and rotational behaviour in 6-hydroxydopamine denervated rats using intracerebral dialysis. Brain Res376:1–7.
+ZhanY, WuH, LiuL, LinJ, ZhangS. 2021. Organic solvent-based tissue clearing techniques and their applications. J Biophotonics14:e202000413.
+ZhaoZ-D, ChenZ, XiangX, HuM, XieH, JiaX, CaiF, CuiY, ChenZ, QianL, LiuJ, ShangC, YangY, NiX, SunW, HuJ, CaoP, LiH, ShenWL. 2019. Zona incerta GABAergic neurons integrate prey-related sensory signals and induce an appetitive drive to promote hunting. Nat Neurosci22:921–932.
+ZweigRM, JankelWR, HedreenJC, MayeuxR, PriceDL. 1989. The pedunculopontine nucleus in Parkinson’s disease. Ann Neurol26:41–46.
+
+
+SUPPLEMENTARY MATERIAL
+
+
+
Time course of open field locomotion distance travelled over 30 minutes.
+
(A-B) 30-minute open field experiment group averages for sham (A) and 6-OHDA (B) animals with photoactivation plotted as 1-minute bins of distance travelled. Blue bars indicate 1-minute trials with photoactivation. (C) Locomotion distance travelled for the six sham ChR2 animals at baseline and at the five pre-timepoints compared using a 1-way RM ANOVA (F5,25 = 0.486, P = .783). Data indicate mean ± SEM bars.
+
+
+
+
+
Characterization of A13 region photoactivation temporal dynamics on locomotion initiation. (A)
+
Percent of trials where there was at least one bout of locomotion. Data are plotted as box and whiskers with the horizontal line through the box indicating the group median, interquartile range indicated by the limits of the box, and group minimum and maximum indicated by the whiskers. (B) The average time for the ChR2 group animals to begin locomotion after the onset of photoactivation. Means plotted with error bars indicating ± SEM. Asterisks indicate significant comparisons using the Wilcoxon signed-rank test: ** P ≤ 0.05.
+
+
+
+
+
Preservation of TH<sup>+</sup> A13 cells in Parkinsonian mouse models.
+
Representative slices of SNc (AP: −3.08 mm, A) and A13 region (AP: −1.355 mm, D) following registration with WholeBrain software 64. Full 3D brain is available (see Movie S4). There was a lack of TH+ SNc cells following 6-OHDA injections at the MFB (A). (B, C) Zoomed sections (90 μm thickness) of red boxes in panel A in left to right order. Meanwhile, TH+ VTA cells were preserved bilaterally. In addition, TH+ A13 cells were present ipsilesional to 6-OHDA injections (D). (E, F) Zoomed sections (90 μm thickness) of red boxes in panel D in left to right order. Scale bars are 50 μm. When calculating the percentage of TH+ cell loss normalized to the intact side, there was a significant interaction between the condition group and brain regions (repeated measures two-way ANOVA with post hoc Bonferroni pairwise, sham: n = 3, 6-OHDA: n = 6) G). 6-OHDA treated mice showed a significantly greater percentage of TH+ cell loss in SNc compared to VTA and A13 region (VTA vs. SNc: P = 0.005; A13 region vs. SNc: P = 0.029). In contrast, sham showed no significant difference in TH+ cell loss across SNc, VTA and A13 region regions (P > 0.05). *P ≤ 0.05, and **P ≤ 0.01. Scale bars are 50 μm unless otherwise indicated.
+
+
+
+
+
Example of the injection core in a sham brain for viral tracers and the rostral and caudal spread to the injection site (A13). Viral tracers (AAV8-CamKII-mCherry and AAVrg-CAG-GFP) were mixed 50:50. Light-sheet images around the injection site were obtained with 2x objective, 6.3x optical zoom, and a z-step size of 2 µm (xyz resolution = 0.477 µm x 0.477 µm x 2 µm). Background filtering (median value of 20 pixels and Gaussian smoothing with a sigma value of 10) was performed in ImageJ software 1 and visualized in IMARIS 9.8 (Belfast, United Kingdom). 2008 Allen reference atlas images were overlaid on top of 90 µm maximum intensity projections taken from IMARIS 9.8 (Belfast, United Kingdom):-1.26 mm (A), −1.36 mm (B), and −1.46 mm (C). Zoomed in sections of each white rectangular area at each coordinate (rows ‘i’) are displayed below for each fluorophore (rows ‘ii’). Scale bars for rows ‘i’ are 200 µm and for rows ‘ii are 100 µm.
+
+
+
+
+
The consistency of afferent and efferent proportions across mice was compared in a pairwise manner. An experimental variation on the total labeling of neurons and fibers was minimized by dividing the afferent cell counts or efferent fiber areas in each brain region by the total number found in a brain to obtain the proportion of total inputs and outputs. Using Spearman’s correlation analysis, we found afferent and efferent proportions across animals to be consistent among each other with an average correlation of 0.91 (SEM = 0.02). M1 = mouse #1, M2 = mouse #2, M3 = mouse #3.
+
+
+
Movie S1. Photoactivation of the A13 region in a 6-OHDA model mouse producing increased locomotion in the OFT (2x speed).
+
Movie S2. Photoactivation of the A13 region in a sham mouse producing increased locomotion in the OFT (2x speed).
+
Movie S3. Photoactivation of the A13 region during the pole test in a 6-OHDA model mouse decreases pole descent time (0.5x speed).
+
Movie S4. Video showing TH staining following whole brain imaging and staining in a 6-OHDA model mouse brain. Focus on TH expression in the A13 and SNc regions.
Diffusional kurtosis imaging (DKI) is a methodology for measuring the extent of non-Gaussian diffusion in biological tissue, which has shown great promise in clinical diagnosis, treatment planning and monitoring of many neurological diseases and disorders. However, robust, fast and accurate estimation of kurtosis from clinically feasible data acquisitions remains a challenge. In this study, we first outline a new accurate approach of estimating mean kurtosis via the sub-diffusion mathematical framework. Crucially, this extension of the conventional DKI overcomes the limitation on the maximum b-value of the latter. Kurtosis and diffusivity can now be simply computed as functions of the sub-diffusion model parameters. Second, we propose a new fast and robust fitting procedure to estimate the sub-diffusion model parameters using two diffusion times without increasing acquisition time as for the conventional DKI. Third, our sub-diffusion based kurtosis mapping method is evaluated using both simulations and the Connectome 1.0 human brain data. Exquisite tissue contrast is achieved even when the diffusion encoded data is collected in only minutes. In summary, our findings suggest robust, fast and accurate estimation of mean kurtosis can be realised within a clinically feasible diffusion weighted magnetic resonance imaging data acquisition time.
+
+
+
+
+
+
+
+
+Introduction
+
Diffusion weighted magnetic resonance imaging (DW-MRI) over a period of more than 30 years has become synonymous with tissue microstructure imaging. Measures of how water diffuses in heterogeneous tissues allow indirect interpretation of changes in tissue microstructure (Le Bihan and Johansen-Berg, 2012). DW-MRI has predominantly been applied in the brain, where properties of white matter connections between brain regions are often studied (Lebel et al., 2019), in addition to mapping tissue microstructural properties (Tournier, 2019). Applications outside of the brain have clinical importance as well, and interest is growing rapidly.
+
Generally, DW-MRI involves the setting of diffusion weightings and direction over which diffusion is measured. While diffusion tensor imaging (DTI) can be performed using a single diffusion weighting, a so-called b-shell, and at least six diffusion encoding directions (Le Bihan et al., 2001), other models tend to require multiple b-shells each having multiple diffusion encoding directions. Diffusional kurtosis imaging (DKI) is a primary example of a multiple b-shell, multiple diffusion encoding direction method (Jensen et al., 2005). DKI is considered as an extension of DTI (Jensen and Helpern, 2010;Hansen et al., 2013; Veraart et al., 2011b), where the diffusion process is assumed to deviate away from standard Brownian motion, and the extent of such deviation is measured via the kurtosis metric. Essentially, the increased sampling achieved via DKI data acquisitions allows more complex models to be applied to data (Van Essen et al., 2013; Shafto et al., 2014), in turn resulting in metrics of increased utility for clinical decision making.
+
Recent clinical benefits of using kurtosis metrics over other DW-MRI derived measures have been demonstrated for grading hepatocellular carcinoma (Li et al., 2022b), prognosing chronic kidney disease (Liu et al., 2021), differentiating parotid gland tumours (Huang et al., 2021a), measuring response to radiotherapy treatment in bone tumour (Guo et al., 2022a) and glioblastoma (Goryawala et al., 2022), identifying tissue abnormalities in temporal lobe epilepsy patients with sleep disorders (Guo et al., 2022b) and brain microstructural changes in mild traumatic brain injury (Wang et al., 2022), monitoring of renal function and interstitial fibrosis (Li et al., 2022a), detecting the invasiveness of bladder cancer into muscle (Li et al., 2022d), aiding management of patients with depression (Maralakunte et al., 2022), delineating acute infarcts with prognostic value (Hu et al., 2022), predicting breast cancer metastasis (Zhou et al., 2022), diagnosing Parkinson’s disease (Li et al., 2022c), amongst others reported prior and not listed here.
+
The routine use of DKI in the clinic has nonetheless lagged due the inability to robustly estimate the kurtosis metric (Veraart et al., 2011a; Tabesh et al., 2010; Kuder et al., 2011; Henriques et al., 2021). A known requirement for estimating kurtosis in DKI is to restrict the maximum b-value to 2000 s/mm2-3000 s/mm2 for brain studies (Jensen et al., 2005; Jensen and Helpern, 2010; Poot et al., 2010), with the optimal maximum b-value found to be dependent on tissue type (Poot et al., 2010). This suggests that the traditional kurtosis model is less accurate at representing the diffusion signal at large b-values. Moreover, multiple b-shell, multiple direction high quality DW-MRI data can take many minutes to acquire, which poses challenges for clinical imaging protocols involving a multitude of MRI contrasts already taking tens of minutes to execute. Reduction of DKI data acquisition times through parallel imaging, optimisation of b-shells and directions have been investigated (Zong et al., 2021; Heidemann et al., 2010; Zelinski et al., 2008), and DW-MRI data necessary for DKI analysis has been shown to supersede the data required for DTI (Veraart et al., 2011b). Therefore, an optimised DKI protocol can potentially replace clinical DTI data acquisitions without adversely affecting the estimation of DTI metrics.
+
For DKI to become a routine clinical tool, DW-MRI data acquisition needs to be fast and provides a robust estimation of kurtosis. The ideal protocol should have a minimum number of b-shells and diffusion encoding directions in each b-shell. The powder averaging over diffusion directions improves the signal-to-noise ratio of the DW-MRI data used for parameter estimation. Whilst this approach loses out on directionality of the kurtosis, it nonetheless provides a robust method of estimating mean kurtosis (Henriques et al., 2021), a metric of significant clinical value.
+
Instead of attempting to improve an existing model-based approach for kurtosis estimation, as has been considered by many others, we considered the problem from a different perspective. In view of the recent generalisation of the various models applicable to DW-MRI data (Yang et al., 2022), the sub-diffusion framework provides new, unexplored opportunities, for fast and robust kurtosis mapping. We report on our investigation into the utility of the sub-diffusion model for practically useful mapping of mean kurtosis.
+
+
+Results
+
Simulation studies were conducted to establish the requirements on the number of diffusion times and the separation between them for accurate estimation of mean kurtosis based on the subdiffusion model augmented with random Gaussian noise, following (10). Testing and validation was performed using the human Connectome 1.0 brain dataset (Tian et al., 2022). The 2 × 2 × 2mm3 resolution data were obtained using two diffusion times (Δ = 19, 49ms) with a pulse duration of δ = 8ms and G = 31,68,105,142,179, 216, 253, 290 mT/m, respectively generating b-values = 50, 350, 800, 1500, 2400, 3450, 4750, 6000 s/mm2, and b-values = 200, 950, 2300, 4250, 6750, 9850, 13500, 17800 s/mm2, according to b-value = (γδG)2(Δ - δ/3). Up to 64 diffusion encoding directions per b-shell were set. The traditional method for mean kurtosis estimation was implemented (producing KDKI), which is limited to the use of DW-MRI generated using a single diffusion time (Jensen et al., 2005; Jensen and Helpern, 2010; Veraart et al., 2011a; Poot et al., 2010), alongside our implementation based on the sub-diffusion model (3), wherein mean kurtosis K* is computed as a function of the sub-diffusion model parameter β (refer to (9)) using either a single or multiple diffusion times.
+
+Multiple diffusion times for robust and accurate mean kurtosis estimation
+
In our simulations we tested up to five distinct diffusion times to generate b-values. Figure 1 illustrates the effects of the number of diffusion times on the parameter estimation at various SNR levels. We draw attention to a number of features in the plots. First, as SNR is increased from 5 to 20 (rows 1-3), the variability in the estimated parameters (Dβ, β, K*) decreases. Second, increasing the number of distinct diffusion times used for parameter estimation decreases estimation variability, with the most significant improvement when increasing from one to two diffusion times (rows 1-3). Third, sampling with two distinct diffusion times provides more robust parameter estimates than sampling twice as many b-values using one diffusion time (compare 2 and 2′ violin plots, rows 1-3). Fourth, the last row (row 4) highlights the improvement in the coefficient of variation (CV) for each parameter estimate with increasing SNR. This result again confirms that the most pronounced decline ofCV occurs when increasing from one to two diffusion times, and parameter estimations can be performed more robustly using DW-MRI data with a relatively high SNR.
+
In Figure 2 we provide simulation results evaluating the choice of the two distinct diffusion times (assuming Δ1 < Δ2) by measuring the goodness-of-fit of the model. The smaller of the two diffusion times is stated along the abscissa, and the difference, i.e., Δ2 - Δ1, is plotted along the ordinate. The quality of fitting was measured using the coefficient of determination (the larger the R2 value, the better the goodness-of-fit of the model) for each combination of abscissa and ordinate values. The conclusion from this figure is that Δ1 should be small, while Δ2 should be as large as practically plausible. Note, DW-MRI echo time was not considered in this simulation, but as Δ2 increases, the echo time has to proportionally increase. Because of the inherent consequence of decreasing SNR with echo time, special attention should be paid to the level of SNR achievable with the use of a specific Δ2. Nonetheless, our findings suggest that when Δ1 = 8 ms, Δ2 can be set as small as 21 ms to achieve an R2 > 0.90 with an SNR as low as 5. If Δ1 is increased past 15 ms, then the separation between Δ1 and Δ2 has to increase as well, and such choices benefit from an increased SNR in the DW-MRI data. The Connectome 1.0 DW-MRI data was obtained using Δ1 = 19 ms and Δ2 = 49 ms, leading to a separation of 30 ms. For this data it is expected that with SNR = 5 an R2 around 0.9 is feasible, and by increasing SNR to 20, the R2 can increase to a value above 0.99.
+
In Figure 3, we present the scatter plots of simulated K values versus fitted K values using simulated data with different number of diffusion times at various SNR levels. Four cases are provided, including fitting simulated data generated with Δ1 = 19 ms (row 1) or Δ2 = 49 ms (row 2), fitting data with both diffusion times (row 3), and fitting data with three diffusion times (row 4). R2 values for each case at each SNR level are provided. This result verifies that sub-diffusion based kurtosis estimation (blue dots) improves using multiple diffusion times. The improvement in R2 is prominent when moving from fitting single diffusion time data to two diffusion times data, especially when the data is noisy (e.g., SNR = 5 and 10). The improvement gained by moving from two to three distinct diffusion times is marginal (less than 0.01 improvement in R2 value at SNR = 5 and 10, and no improvements for SNR = 20 data). Moreover, our simulation findings highlight the deviation away from the ground truth kurtosis K by using the traditional DKI method (orange dots), especially with kurtosis values larger than 1. Overall, fitting sub-diffusion model (3) to data with two adequately separated diffusion times can lead to robust estimation of mean kurtosis, via (9).
+
+
+Time-dependence in DKI metrics
+
In Figure 4, we show the time-dependence effect of the DKI metrics (DDKI and KDKI) after fitting the standard DKI model to our simulated data. In this fitting, we consider b-values of0, 1000, 1400 and 2500, and vary the diffusion time, as in Jelescu et al. (2022). We depict the parameter estimates, DDKI and KDKI, from simulated data with added noise (SNR = 20) in Figure 4(A) and (B) for the diffusion time between 10-110ms. In Figure 4 (C) and (D), using data with no added noise, we illustrate the long term fitting results and trends in the parameter estimates. In both (A) and (C), as diffusion time increases, DDKI decreases as expected for an effective diffusion coefficient. The mean DDKI (averaged over 1000 simulations) agrees with the ground truth diffusivity D*. When it comes to the kurtosis KDKI, in the noiseless data setting (D), we see KDKI is converging to the ground truth kurtosis value K* at large diffusion time, while in the noisy data setting (B), this trend is not obvious within the experimental diffusion time window. More explanations of the observed time-dependence of diffusivity and kurtosis are provided in the Discussion.
+
+
+
Simulation results on the effect of the number of diffusion times involved in the fitting of the sub-diffusion model (3) parameters (Dβ, β) and computing K* following (9) at various SNR levels. The ground truth values for (Dβ, β, K*) are set to (3 × 10-4, 0.75, 0.8125) to represent white matter (blue) and (5 × 10-4, 0.85, 0.4733) to represent gray matter (orange). Rows 1-3: Distributions of fitted parameter values using different number of diffusion times. 2′ represents an additional simulation using two diffusion times but set to be the same, so it has the same number of data points in the fitting as for using two different diffusion times. Row 4: Coefficient of variation (CV) of the parameter values fitted using different number of diffusion times.
+
+
+
+
+
Surface plots of R2 values achieved with fitting simulated data with two diffusion times, Δ1 and Δ2, to the sub-diffusion model (3) at various SNR levels. R2 values were computed by comparing the estimated mean kurtosis with the ground truth kurtosis. R2 contours at the 0.85, 0.90, 0.95 and 0.99 levels have been provided for visualisation purposes.
+
+
+
+
+Towards fast DKI data acquisitions
+
Next, we sought to identify the minimum number and combination of b-values to use for mean kurtosis estimation based on the sub-diffusion model (3). The simulation results were generated using the Connectome 1.0 DW-MRI data b-value setting, which has 16 b-values, 8 from Δ = 19ms and 8 from Δ = 49ms.
+
In Figure 5, we computed R2 values for all combinations of choices when the number of nonzero b-values is two, three, and four. We then plotted the R2 values sorted in ascending order for each considered SNR level. Notably, as the number of non-zero b-values increase, the achievable combinations increase as well (i.e., 120, 560 and 1820 for the three different non-zero b-value cases). The colours illustrate the proportion of b-values used in the fitting based on Δ1 and Δ2. For example, 1 : 0 means only Δ1 b-values were used, and 2 : 1 means two Δ1 and one Δ2 b-values were used to generate the result. The b-value combinations achieving highest R2 values are displayed in the inset pictures and the b-values are provided in Table 1.
+
Our simulation results in Table 1 suggest that increasing the number of non-zero b-values from two to four improves the quality of the parameter estimation, also achievable by fitting the subdiffusion model (3) to higher SNR data. The gain is larger by using higher SNR data than by using more b-values. For example, going from two to four non-zero b-values with SNR = 5 data approximately doubles the R2, whereas the R2 almost triples when SNR = 5 data is substituted by SNR = 20 data. Additionally, the use of Δ1 or Δ2 alone is not preferred (also see Figure 5), and preference is towards first using Δ1 and then supplementing with b-values generated using Δ2. At all SNR levels when only two non-zero b-values are used, one b-value should be chosen based on the Δ1 set, and the other based on Δ2. Moving to three non-zero b-values requires the addition of another Δ1 b-value, and when four non-zero b-values are used then two from each diffusion time are required. If we consider an R2 = 0.90 to be a reasonable goodness-of-fit for the sub-diffusion model, then at least three or four non-zero b-values are needed with an SNR = 20. If SNR = 10, then three non-zero b-values will not suffice. Interestingly, an R2 of 0.85 can still be achieved when SNR = 20 and two optimally chosen non-zero b-values are used.
+
+
+
Scatter plots of simulated K values vs. fitted K values for simulated data with different number of diffusion times at various SNR levels. The simulated data is created using the sub-diffusion model with random normal noise (10). Blue dots represent kurtosis based on fitting the sub-diffusion model (3). Orange dots represent kurtosis based on fitting the traditional DKI model (6). Black line is a reference line for R2 = 1.00, indicating fitted kurtosis values are 100% matching the simulated ones.
+
+
+
+
+
Time-dependence in DKI metrics using simulated data at different diffusion times (
+
+). The ground truth values for (Dβ, β, K*) used in the simulations are set to (3 × 10-4, 0.75, 0.8125) to represent white matter (blue dotted lines) and (5 × 10-4, 0.85, 0.4733) to represent gray matter (orange dotted lines). (A) and (B) use data with added random Gaussian noise (SNR = 20) to estimate the parameters DDKI and KDKI. (C) and (D) use noiseless data to obtain estimates for large
+
+ values. Shaded regions in (A) and (B) represent the 95% confidence intervals of the estimates.
+
+
+
+
+
R2 values for the b-value sampling optimisation based on DW-MRI data with SNR = 5, 10 and 20. The specifically investigated b-value combinations using two, three and four non-zero b-values have been ordered by the size of the R2 value. The colour bar depicts the proportion of Δ = 19 ms and Δ = 49 ms b-values needed to produce the corresponding R2 value. Note, the different b-value combinations were assigned a unique identifier, and these appear along the abscissa for each of the three non-zero b-value cases. The b-value combinations achieving the highest R2 values are displayed in the inset pictures and the b-values are provided in Table 1.
+
+
+
+
+
A selection of the best b-value sampling regimes to achieve the highest R2 value in the three cases considered. The various categories correspond with two, three and four non-zero b-value sampling schemes, with Δ1 and Δ2 denoting the diffusion time setting used to generate the b-values. Note, entries are b-values in unit of s/mm2, and Δ1 = 19 ms and Δ2 = 49 ms were used to match the Connectome 1.0 DW-MRI data collection protocol. The entries listed at the bottom row are suggested optimal nonzero b-values for clinical practice.
+
+
+
Table 1 summarises findings based on having different number of non-zero b-values with R2 values deduced from the SNR = 5, 10 and 20 simulations. We have chosen to depict five b-value combinations producing the largest R2 values for the two, three and four non-zero b-value sampling cases. We found consistency in b-value combinations across SNR levels. Thus, we can conclude that a range of b-values can be used to achieve a large R2 value, which is a positive finding, since parameter estimation does not stringently rely on b-value sampling. For example, using three non-zero b-values an R2 ≥ 0.90 is achievable based on different b-value sampling. Importantly, two distinct diffusion times are required, and preference is towards including a smaller diffusion time b-value first. Hence, for three non-zero b-values we find two b-values based on Δ1 and one based on Δ2. This finding suggests one of the Δ1 b-values can be chosen in the range 50 s/mm2 to 350 s/mm2, and the other in the range 1500 s/mm2 to 4750 s/mm2. Additionally, the Δ2 b-value can also be chosen in a range, considering between 2300 s/mm2 to 4250 s/mm2 based on the Connectome 1.0 b-value settings. The b-value sampling choices made should nonetheless be in view of the required R2 value. Overall, sampling using two distinct diffusion times appears to provide quite a range of options for the DW-MRI data to be used to fit the sub-diffusion model parameters. The suggested optimal b-value sampling in the last row of Table 1, primarily chosen to minimise b-value size whilst maintaining a large R2 value, may be of use for specific neuroimaging studies, which will be used to inform our discussion on feasibility for clinical practice.
+
+
+Benchmark mean kurtosis in the brain
+
The benchmark mean kurtosis estimation in the brain is established using the entire b-value range with all diffusion encoding directions available in the Connectome 1.0 DW-MRI data. For two subjects in different slices, Figure 6 provides the spatially resolved maps of mean kurtosis computed using the sub-diffusion method (i.e., K*) with one or two diffusion times, and using the standard method (i.e., KDKI) considering the two distinct diffusion times. First, we notice a degradation in the KDKI image with an increase in diffusion time. Second, the use of a single diffusion time with the sub-diffusion model leads to K* values which are larger than either the KDKI values or K* values generated using two diffusion times. Third, the quality of the mean kurtosis map appears to visually be best when two diffusion times are used to estimate K*. Superior grey-white matter tissue contrast (TC) was found for the K* map (TC = 1.73), compared to the KDKI maps (TC = 0.80 for the Δ = 19 ms dataset and TC = 1.01 for the Δ = 49 ms dataset).
+
In Figure 7, an error map (measured by root-mean-square-error, RMSE) from Subject 5 slice 74 was presented for fitting the sub-diffusion model to the DW-MRI data with two diffusion times. Sample parameter fittings in both b-space (3) and q-space (4) were provided for four representative white and grey matter voxels.
+
Quantitative findings are provided in Table 2. The analysis was performed for sub-cortical grey matter (scGM), cortical grey mater (cGM) and white matter (WM) brain regions. For specifics we refer the reader to the appropriate methods section. The table entries highlight the differences in mean kurtosis when computed using the two different approaches. The trend for the traditional single diffusion time approach is that an increase in Δ results in a slight decrease in the mean KDKI, and an increase in the coefficient of variation (CV) for any region. For example, the mean KDKI in the thalamus reduces from 0.65 to 0.58, while the CV increases from 30% to 39%. As much as 30% increase in CV is common for scGM and cGM regions, and around 10% for WM regions. The CV based on the K* value for each region is less than the CV for KDKI with either Δ = 19 ms or Δ = 49 ms.
+
Figure 8 presents the distributions of the fitted parameters (D and K) in specific brain regions, based on the sub-diffusion model (panel A) and the standard DKI model with Δ = 19 ms (panel B). The distributions are colored by the probability density. Yellow indicates high probability density, light blue indicates low probability density. In each subplot, the diffusivity is plotted along the abscissa axis and the kurtosis is along the ordinate axis. Results for the standard DKI model with Δ = 49 ms are qualitatively similar, so are not shown here. From panel (B), we see an unknown nonlinear relationship between the DKI pair, DDKI and KDKI, in all regions considered. By comparison, the sub-diffusion based K* and D* (panel A) are less correlated with each other, indicating D* and K* carry distinct information, which will be very valuable for characterising tissue microstructure.
+
+
+Reduction in DKI data acquisition
+
Results for reductions in diffusion encoding directions to achieve different levels of SNR with the purpose of shortening acquisition times will be benchmarked against the K* maps in Figure 6 and the K* values reported in Table 2.
+
Figure 9 presents the qualitative findings for two subjects generated using all, optimal and sub-optimal b-value samplings with SNR = 6 (3 non-collinear directions, 6 measurements), 10 (8 non-collinear directions, 16 measurements) and 20 (32 non-collinear directions, 64 measurements) DW-MRI data. The quality of the mean kurtosis map improves with increasing SNR, and also by optimising b-value sampling. Optimal sampling at SNR = 10 is qualitatively comparable to the SNR = 20 optimal sampling result, and to the benchmark sub-diffusion results in Figure 6.
+
Quantitative verification of the qualitative observations are provided in Table 3. Significant differences in brain region-specific mean kurtosis values occur at the SNR = 6 level, which are not apparent when SNR = 10 or 20 data with optimal b-value sampling were used. The average errors are relative errors compared to the benchmark kurtosis values reported in Table 2. When using optimal b-values, average errors range from 22% to 43% at SNR = 6, from 13% to 43% at SNR = 10, and from 8% to 20% at SNR = 20, across brain regions. When using sub-optimal b-values, average errors range from 47% to 57% at SNR = 6, from 24% to 102% at SNR = 10, and from 27% to 72% at SNR = 20. Also, the brain region-specific CV for mean kurtosis was not found to change markedly when SNR = 10 or 20 data were used to compute K*. The result of reducing the SNR to 6 leads to an approximate doubling of the CV for each brain region. These findings confirm that with optimal b-value sampling, the data quality can be reduced to around the SNR = 10 level, without a significant impact on the region-specific mean kurtosis estimates derived using the sub-diffusion model.
+
+
+
Spatially resolved maps of mean kurtosis shown for two example slices and two different subjects, Subject 3 rescan slice 71 (Panel A) and Subject 5 slice 74 (Panel B) from the Connectome 1.0 DW-MRI data. Individual maps were generated using the sub-diffusion model framework (K*), as well as using the traditional approach (KDKI). The diffusion times, Δ, used to generate each plot are provided for each case. We consider the mean kurtosis maps using two diffusion times (Δ = 19, 49ms) as the benchmarks.
+
+
+
+
+
Representative error map and sample parameter fits for Subject 5 slice 74. The DW-MRI data with two diffusion times was fitted to the sub-diffusion model in both q-space (A-D) and b-space (E-H), following (3) and (4) respectively. The first and second columns are voxels in white matter (30,20,74) and (45,56,74), respectively. The third and fourth columns are voxels in grey matter (58,35,74) and (34,78,74), respectively.
+
+
+
+
+
Benchmark kurtosis values estimated using the Connectome 1.0 DW-MRI data for different regions of the human brain. Results are provided for the traditional mean kurtosis (KDKI) at two distinct diffusion times, and values (K*) obtained based on fitting the sub-diffusion model across both diffusion times. Results are for grey matter (GM) and white matter (WM) brain regions, in categories of sub-cortical (sc) and cortical (c), and CC stands for corpus callosum. The pooled means and standard deviations across participants have been tabulated, along with the coefficient of variation in parentheses.
+
+
+
+
+
Distributions of the estimated parameter pair (D,K) in different regions of the brain of all subjects, colored by the probability density. Yellow indicates high probability density, light blue indicates low probability density. Panel A: distributions of (D*,K*), generated using the sub-diffusion model (3) with both Δ = 19, 49ms. Panel B: distributions of (DDKI,KDKI), generated using the standard DKI model (6) with Δ = 19ms. Kurtosis is dimensionless and diffusivity is in units of × 10-3 mm2/s.
+
+
+
+
+
Spatially resolved maps of mean kurtosis shown for two example slices and two different subjects, Subject 3 rescan slice 71 (Panel A) and Subject 5 slice 74 (Panel B), based on SNR reduction of the Connectome 1.0 DW-MRI data. Individual maps were generated using the sub-diffusion model framework (K*), considering optimal and sub-optimal four non-zero b-value sampling schemes. Here, two b-values with Δ = 19 ms and two b-values with Δ = 49 ms were selected for each case. The optimal b-values were chosen as the best for each SNR shown in Table 1. The sub-optimal b-values were chosen to have an R2 = 0.3, 0.45, 0.5 to be about half of the maximum R2, for SNR = 6 (b = 800, 1500,200,2300 s/mm2), SNR = 10 (b = 1500, 3450, 6750, 13500 s/mm2) and SNR = 20 (b = 3450, 4750, 2300, 4250 s/mm2), respectively. The benchmark kurtosis map is provided in Figure 6.
+
+
+
+
+Scan-rescan reproducibility of mean kurtosis
+
Figure 10 summarises the intraclass correlation coefficient (ICC) distribution results (μ for mean; σ for standard deviation) for the specific brain regions analysed. The two sets of ICC values were computed based on all DW-MRI (i.e., SNR = 20; subscript A) and the SNR = 10 optimal b-value sampling scheme (subscript O). As the value of μ approaches 1, the inter-subject variation in mean kurtosis is expected to greatly outweigh the intra-subject scan-rescan error. The value of μ should always be above 0.5, otherwise parameter estimation cannot be performed robustly and accurately, and values above 0.75 are generally accepted as good. The μA values for all regions were in the range 0.76 (thalamus) to 0.87 (caudate), and reduced to the range 0.57 (thalamus) to 0.80 (CC) when optimal sampling with SNR = 10 was used to estimate the K* value. Irrespective of which of the two DW-MRI data were used for K* estimation, the value of p. was greater than or equal to 0.70 in 20 out of 24 cases. The μO was less than 0.70 for only the thalamus, putamen and pallidum. The loss in ICC by going to SNR = 10 data with optimal b-value sampling went hand-in-hand with an increase in σ, which is not unexpected, since the uncertainty associated with using less data should be measurable. Overall, μA, μO, and σA, σO, were fairly consistent across the brain regions, suggesting the DW-MRI data with SNR = 10 is sufficient for mean kurtosis estimation based on the sub-diffusion framework.
+
+
+
+Discussion
+
DW-MRI allows the measurement of mean kurtosis, a metric for the deviation away from standard Brownian motion of water in tissue, which has been used to infer variations in tissue microstructure. Research on mean kurtosis has shown benefits in specific applications over other diffusion related measures derived from DW-MRI data (Li et al., 2022b; Liu et al., 2021; Huang et al., 2021a; Guo et al., 2022a; Goryawala et al., 2022; Guo et al., 2022b; Wang et al., 2022; Li et al., 2022a,d; Maralakunte et al., 2022; Hu et al., 2022; Zhou et al., 2022; Li et al., 2022c). Whilst many efforts have been made to optimise mean kurtosis imaging for clinical use, the limitations have been associated with lack of robustness and the time needed to acquire the DW-MRI data for mean kurtosis estimation. The choice of the biophysical model and how diffusion encoding is applied are critical to how well kurtosis in the brain is mapped. Here, we evaluated the mapping of mean kurtosis based on the sub-diffusion model, which allows different diffusion times to be incorporated into the data acquisition. Using simulations and the Connectome 1.0 public DW-MRI dataset, involving a range of diffusion encodings, we showed that mean kurtosis can be mapped robustly and rapidly provided at least two different diffusion times are used and care is taken towards how b-values are chosen given differences in the SNR level of different DW-MRI acquisitions.
+
+Reduction in scan time
+
Previous attempts have been made in optimising the DW-MRI acquisition protocol for mean kurtosis estimation based on the traditional, single diffusion time kurtosis model (Hansen et al., 2013; Hu et al., 2022; Poot et al., 2010; Hansen et al., 2016). Considerations have been made towards reducing the number of b-shells, directions per shell, and sub-sampling of DW-MRI for each direction in each shell. Our findings suggest that robust estimation of kurtosis cannot be achieved using the classical model for mean kurtosis, as highlighted previously (Yang et al., 2022; Ingo et al., 2014, 2015; Barrick et al., 2020). A primary limitation of the traditional method is the use of the cumulant expansion resulting in sampling below a b-value of around 2500 s/mm2 (Jensen etal., 2005; Jensen and Helpern, 2010), and using the sub-diffusion model this limitation is removed (Yang et al., 2022). Our simulation findings and experiments using the Connectome 1.0 data confirm that mean kurtosis can be mapped robustly and rapidly using the sub-diffusion model applied to an optimised DW-MRI protocol. Optimisation of the data acquisition with the use of the sub-diffusion model has not been considered previously.
+
+
+
Kurtosis values (K*) under the optimal and sub-optimal b-value sampling regimes for specific brain regions. K* was estimated based on fitting the sub-diffusion model to the Connectome 1.0 DW-MRI data with two diffusion times and selected four b-shells. Optimal b-value sampling is considered to have R2 = 0.63, 0.91 and 0.96 for the SNR = 6, 10 and 20 columns, according to Table 1. Sub-optimal b-values are chosen to have R2 = 0.3, 0.45 and 0.5, respectively, as reported in Figure 9. Individual entries are for grey matter (GM) and white matter (WM) brain regions, in categories of sub-cortical (sc) and cortical (c), and CC stands for corpus callosum. A reduction in SNR level was achieved by reducing the number of diffusion encoding directions in each b-shell of the DW-MRI data. The pooled means and standard deviations across participants have been tabulated, along with the coefficient of variation in parentheses. The entries identified in italic under the optimal b-value heading were found to be significantly different from the mean K* reported in Table 2. Sub-optimal result population means were mostly significantly different from the mean K*, and they are not italicised. The average errors (last column) are relative errors compared to the benchmark kurtosis values reported in Table 2.
+
+
+
+
+
Interclass correlation coefficient (ICC) results for mean kurtosis are depicted for the 12 brain regions analysed. The mean (μ) and standard deviation (σ) computed based on all the Connectome 1.0 DW-MRI data (A), and the reduced data achieving an SNR = 10 with optimal four non-zero b-value sampling (O), are provided for each brain region. Histograms were generated using all data. Mean kurtosis based on the optimised protocol was computed using the sub-diffusion framework using DW-MRI data with the four non-zero b-values suggested in Table 1 and diffusion encoding directions down sampled to achieve an SNR = 10.
+
+
+
Mean kurtosis values can be generated based on having limited number of diffusion encoding directions (refer to Figure 9 and Table 3). Given that each direction for each b-shell takes a fixed amount of time, then a four b-shell acquisition with six directions per shell will take 25 times longer than a single diffusion encoding data acquisition (assuming a single b-value = 0 data is collected). The total acquisition time for the diffusion MRI protocol for the Connectome 1.0 data was 55 minutes, including 50 b = 0 s/mm2 scans plus seven b-values with 32 and nine with 64 diffusion encoding directions (Tian et al., 2022). This gives a total of 850 scans per dataset. As such, a single 3D image volume took 3.88 s to acquire. Conservatively allowing 4 s per scan, and considering SNR = 20 data (i.e. 64 directions) over four b-values and a single b-value = 0 scan, DW-MRI data for mean kurtosis estimation can be completed in 17 min 8 s (R2 = 0.96). At SNR = 10 (i.e. 32 directions), DW-MRI data with the same number of b-values can be acquired in 4 min 20 s (R2 = 0.91). If an R2 = 0.85 (SNR = 10) is deemed adequate, then one less b-shell is required, saving an additional 64 s. We should point out that even though 2-shell optimised protocols can achieve R2 = 0.85 with SNR = 20, this is not equivalent in time to using 3-shells with SNR = 10 (also R2 = 0.85). This is because 4x additional data are required to double the SNR (equivalent to acquiring an additional 4-shells). However, only one extra b-shell is required to convert 2-shell data to 3-shells with SNR = 10. Our expected DW-MRI data acquisition times are highly feasible clinically, where generally neuroimaging scans take around 15 min involving numerous different MRI contrasts and often a DTI acquisition.
+
An early study on estimating mean kurtosis demonstrated the mapping of a related metric in less than 1 min over the brain (Hansen et al., 2013). Clinical adoption of the protocol lacked, possibly since b-values are a function δ, G and Δ. Hence, different b-values can be obtained using different DW-MRI protocol settings, leading to differences in the mapping of mean kurtosis based on the data (we showed the Δ effect in Figure 6 and Table 2). Our findings suggest this impediment is removed by sampling and fitting data with b-values across two distinct diffusion times. Nonetheless, we should consider what might be an acceptable DW-MRI data acquisition time.
+
A recent study on nasopharyngeal carcinoma investigated reducing the number of b-shell signals based on fixing diffusion encoding directions to the three Cartesian orientations (He et al., 2022). The 3-shell acquisitions took 3 min 2 s to acquire, while the 5-shell data required 5 min 13 s. They investigated as well the impact of using partial Fourier sampling, i.e., reducing the amount of data needed for image reconstruction by reducing k-space line acquisitions for each diffusion encoded image. Their benchmark used 5-shells (200, 400, 800, 1500, 2000 s/mm2), and found partial Fourier sampling with omission of the 1500 s/mm2 b-shell produced acceptable results. With this acquisition the scan could be completed in 3 min 46 s, more than 2× faster than the benchmark 8 min 31 s. Our proposed 3-shells acquisition with an R2 = 0.85 (see SNR = 10 results in Table 1) executable under 4 min is therefore inline with current expectations. Note, at the R2 = 0.85 level the ICC for the different brain regions were in the range 0.60 to 0.69, and these were not formally reported in Figure 10. This level of reproducibility is still acceptable for routine use. We should additionally point out that we used the Subject 1 segmentation labels, after having registered each DW-MRI data to the Subject 1 first scan. This approach results in slight mismatch of the regionspecific segmentations when carried across subjects, inherently resulting in an underestimation of ICC values.
+
Less than 4 min DW-MRI data acquisitions can potentially replace existing data acquisitions used to obtained DTI metrics, since even the estimation of the apparent diffusion coefficient improves by using DW-MRI data relevant to DKI (Veraart et al., 2011b; Wu and Cheung, 2010). Additionally, it is increasingly clear that in certain applications the DKI analysis offers a more comprehensive approach for tissue microstructure analysis (Li et al., 2022b; Liu et al., 2021; Huang et al., 2021a;Guo et al., 2022a; Goryawala et al., 2022; Guo et al., 2022b; Wang et al., 2022; Li et al., 2022a,d; Maralakunte et al., 2022; Hu et al., 2022; Zhou et al., 2022; Li et al., 2022c). As such, multiple b-shell, multiple diffusion encoding direction DW-MRI acquisitions should be used for the calculation of both DTI and DKI metrics.
+
+
+DW-MRI data acquisition considerations
+
To achieve R2 > 0.92 for estimating kurtosis K*, it is necessary to have four b-values, e.g., two relatively small b-values (350 s/mm2 using Δ = 19 ms, and 950 s/mm2 using Δ = 49 ms, both with G = 68 mT/m) and two larger b-values of around 1500 s/mm2 and 4250 s/mm2 (using Δ = 19 ms and Δ = 49 ms respectively, both with G = 142 mT/m) (see bottom row in Table 1). If two or three non-zero b-values are considered sufficient (with R2 = 0.85 or 0.90), then the larger Δ needs to be used to set the largest b-value to be 2300 s/mm2, and the other(s) should be set using the smaller Δ. For the two non-zero b-values case, the b-value from the smaller Δ should be around 800 s/mm2. For the three non-zero b-values case, the b-values from the smaller Δ would then need to be 350 s/mm2 and 1500 s/mm2. Interestingly, b-value = 800 s/mm2 lies around the middle of the 350 s/mm2 to 1500 s/mm2 range. The additional gain to R2 = 0.92 can be achieved by splitting the b-value with the larger Δ into two, again with 2300 s/mm2 near the middle of the two new b-values set. In addition, the separation between Δ1 and Δ2 needs to be as large as plausible, as can be deduced from the simulation result in Figure 2, but attention should be paid to signal-to-noise ratio decreases with increased echo times (He et al., 2022).
+
A recent study on optimising quasi-diffusion imaging (QDI) considered b-values up to 5000 s/mm2 (Spilling et al., 2022). While QDI is a non-Gaussian approach, it is different to the sub-diffusion model, but still uses the Mittag-Leffler function and involves the same number of model parameters. The DW-MRI data used in their study was acquired with a single diffusion time. Nevertheless, particular points are worth noting. Their primary finding was parameter dependence on the maximum b-value used to create the DW-MRI data. They also showed the accuracy, precision and reliability of parameter estimation are improved with increased number of b-shells. They suggested a maximum b-value of 3960 s/mm2 for the 4-shell parameter estimation. Our results do not suggest a dependence of the parameter estimates on maximum b-value (note, if a maximum b-value dependence is present, benchmark versus optimal region specific results in Tables 2 and 3 should show some systematic difference; and we used the real part of the DW-MRI data and not the magnitude as commonly used). These findings potentially confirm that Δ separation is an important component of obtaining a quality parameter fit from which mean kurtosis is deduced (Figure 2).
+
+
+Diffusion gradient pulse amplitudes
+
Commonly available human clinical MRI scanners are capable of 40 mT/m gradient amplitudes. Recently, the increased need to deduce tissue microstructure metrics from DW-MRI measurements has led to hardware developments resulting in 80 mT/m gradient strength MRI scanners. These were initially sought by research centres. The Connectome 1.0 scanners achieve 300 mT/m gradient amplitudes (Tian et al., 2022), which in turn allow large b-values within reasonable echo times. The Connectome 2.0 scanner is planned to achieve 500 mT/m gradient amplitudes (Huang et al., 2021b), providing data for further exploration of the (q, Δ) space by providing a mechanism for increasing our knowledge of the relationships between the micro-, meso- and macro-scales. Whilst there are three Connectome 1.0 scanners available, only one Connectome 2.0 scanner is planned for production. Hence, robust and fast methods utilising existing 80 mT/m gradient systems are needed, and the Connectome scanners provide opportunities for testing and validating methods.
+
The b-value is an intricate combination of δ, G, and Δ. An increase in any of these three parameters increases the b-value (note, δ and G increase b-value quadratically, and Δ linearly). An increase in δ can likely require an increase in Δ, the consequence of which is a reduction in signal-to-noise ratio as the echo time has to be adjusted proportionally. Partial Fourier sampling methods aim to counteract the need to increase the echo time by sub-sampling the DW-MRI data used to generate an image for each diffusion encoding direction (Zong et al., 2021; Heidemann et al., 2010). The ideal scenario, therefore, is to increase G, as for the Connectome 1.0 and 2.0 scanners.
+
Based on our suggested b-value settings in Table 1, a maximum b-value of around 4250 s/mm2 is required for robust mean kurtosis estimation (assume R2 > 0.90 is adequate and achieved via four non-zero b-values and two distinct As, and we should highlight that b-value = 1500 s/mm2 (Δ = 19 ms) and b-value = 4250 s/mm2 (Δ = 49 ms) were achieved using G = 142 mT/m). Considering 80 mT/m gradient sets are the new standard for MRI scanners, an adjustment to δ to compensate for G is needed (recall, q = γδG). Hence, for a constant q, G can be reduced at the consequence of proportionally increasing δ. This then allows MRI systems with lower gradient amplitudes to generate the relevant b-values. For example, changing the δ from 8 ms to 16 ms would result in halving of G (using a maximum G of 142 mT/m from the Connectome 1.0 can achieve the optimal b-values; hence halving would require around 71 mT/m gradient pulse amplitudes). Increasing Δ above 49 ms to create larger b-value data is unlikely to be a viable solution due to longer echo times leading to a loss in SNR. SNR increases afforded by moving from 3 Tto 7 T MRI are most likely counteracted by an almost proportional decrease in T2 times (Pohmann et al., 2016), in addition to 7 T MRI bringing new challenges in terms of increased transmit and static field inhomogeneities leading to signal inconsistencies across an image (Kraff and Quick, 2017).
+
+
+Relationship between sub-diffusion based mean kurtosis <italic>K</italic><sup>*</sup> and histology
+
Following Table 2 (last column), white matter regions showed high kurtosis (0.87±0.22), consistent with a structured heterogeneous environment comprising parallel neuronal fibres, as shown in Maiter et al. (2021). Cortical grey matter showed low kurtosis (0.40±0.16). Subcortical grey matter regions showed intermediate kurtosis (0.60±0.21). In particular, caudate and putamen showed similar kurtosis to grey matter, while thalamus and pallidum showed similar kurtosis properties to white matter. Histological staining results of the subcortical nuclei (Maiter et al., 2021) showed the subcortical grey matter was permeated by small white matter bundles, which could account for the similar kurtosis in thalamus and pallidum to white matter. These results confirmed that our proposed sub-diffusion based mean kurtosis K* is consistent with published histology of normal human brain.
+
+
+Time-dependence of diffusivity and kurtosis
+
The time-dependence of diffusivity and kurtosis has attracted much interest in the field of tissue microstructure imaging. Although our motivation here is not to map time-dependent diffusion, we can nonetheless point out that the assumption ofa sub-diffusion model provides an explanation of the observed time-dependence of diffusivity and kurtosis. From (5), the diffusivity that arises from the sub-diffusion model is of the form
+
+, where
+
+ is the effective diffusion time, DSUB has the standard units for a diffusion coefficient mm2/s, and Dβ is the anomalous diffusion coefficient (with units mm2/sβ) associated with sub-diffusion. In the sub-diffusion framework (1), Dβ and β are assumed to be constant, and hence DSUB exhibits a fractional power-law dependence on diffusion time. Then, following (8), D* is obtained simply by scaling DSUB by a constant 1/Γ(1 + β), and hence also follows a fractional power-law dependence on diffusion time. This time-dependence effect of diffusivity was illustrated in our simulation results, Figure 4(A) and (C).
+
When it comes to kurtosis, the literature on the time-dependence is mixed. Some work showed kurtosis to be increasing with diffusion time in both white and gray matter (Aggarwal et al., 2020), and in gray matter (Ianus et al., 2021), while others showed kurtosis to be decreasing with diffusion time in gray matter (Lee et al., 2020; Olesen et al., 2022; Jelescu et al., 2022). In this study, we provide an explanation of these mixed results. We construct a simulation of diffusion MRI signal data based on the sub-diffusion model (3) augmented with random Gaussian noise. Then we fit the conventional DKI model to the synthetic data. As shown in Figure 4(D), when there is no noise, KDKI increases with diffusion time in white matter, while decreasing with diffusion time in gray matter. When there is added noise, as shown in Figure 4(B), the time-dependency of kurtosis within the timescale of a usual MR experiment is not clear. This goes some way to explaining why the results in the literature on the time-dependence of kurtosis are quite mixed.
+
Furthermore, we summarise the benefits of using the sub-diffusion based mean kurtosis measurement K*. First, as shown in (9), sub-diffusion based mean kurtosis K* is not time-dependent, and hence has the potential to become a tissue-specific imaging biomarker. Second, the fitting of the sub-diffusion model is straightforward, fast and robust, from which the kurtosis K* is simply computed as a function of the sub-diffusion model parameter β, (9). Third, the kurtosis K* is not subject to any restriction on the maximum b-value, as in standard DKI. Hence its value truly reflects the information contained in the full range of b-values.
+
+
+Extension to directional kurtosis
+
The direct link between the sub-diffusion model parameter β and mean kurtosis is well established (Yang et al., 2022; Ingo et al., 2014, 2015). An important aspect to consider is whether mean β used to compute the mean kurtosis is alone sufficient for clinical decision making. While benefits of using kurtosis metrics over other DW-MRI data derived metrics in certain applications are clear, the adequacy of mean kurtosis over axial and radial kurtosis is less apparent. Most studies perform the mapping of mean kurtosis, probably because the DW-MRI data can be acquired in practically feasible times. Nonetheless, we can point to a few recent examples where the measurement of directional kurtosis has clear advantages. A study on mapping tumour response to radiotherapy treatment found axial kurtosis to provide the best sensitivity to treatment response (Goryawala et al., 2022). In a different study a correlation was found between glomerular filtration rate and axial kurtosis is assessing renal function and interstitial fibrosis (Li et al., 2022a). Uniplor depression subjects have been shown to have brain region specific increases in mean and radial kurtosis, while for bipolar depression subjects axial kurtosis decreased in specific brain regions and decreases in radial kurtosis were found in other regions (Maralakunte et al., 2022). This selection of studies highlight future opportunities for extending the methods to additionally map axial and radial kurtosis.
+
Notably, estimates for axial and radial kurtosis require directionality of kurtosis to be resolved, resulting in DW-MRI sampling over a large number of diffusion encoding directions within each b-shell (Jensen and Helpern, 2010; Poot et al., 2010). As such, extension to directional kurtosis requires a larger DW-MRI dataset acquired using an increased number of diffusion encoding directions. The number of b-shells and directions therein necessary for robust and accurate mapping of directional kurtosis based on the sub-diffusion model is an open question.
+
There are three primary ways of determining mean kurtosis. These include the powder averaging over diffusion encoding directions in each shell, and then fitting the model, as in our approach. A different approach is to ensure each b-shell in the DW-MRI data contains the same diffusion encoding directions, and then kurtosis can be estimated for each diffusion encoding direction, after which the average over directions is used to state mean kurtosis. Lastly, the rank-4 kurtosis tensor is estimated from the DW-MRI data, from which mean kurtosis is computed directly. The latter two approaches are potential candidates for extending to axial and radial kurtosis mapping. Note, in DTI a rank-2 diffusion tensor with six unique tensor entries is needed to be estimated, whilst in DKI in addition to the rank-2 diffusion tensor, the kurtosis tensor is rank-4 with 15 unknowns, resulting in 21 unknowns altogether (Hansen et al., 2016). As such, DKI analysis for directional kurtosis requires much greater number of diffusion encoding directions to be sampled than DTI. This automatically means that at least 22 DW-MRI data (including b-value = 0) with different diffusion encoding properties have to be acquired (Jensen and Helpern, 2010). The traditional approach has been to set five distinct b-values with 30 diffusion encoding directions within each b-shell (Poot et al., 2010). Hence, to obtain the entries of the rank-4 kurtosis tensor, much more DW-MRI data is needed in comparison to what is proposed for mean kurtosis estimation in this study. Estimation of the tensor entries from this much data is prone to noise, motion and image artifacts in general (Tabesh et al., 2010), posing challenges on top of long DW-MRI data acquisition times.
+
A kurtosis tensor framework based on the sub-diffusion model where separate diffusion encoding directions are used to fit a direction specific β is potentially an interesting line of investigation for the future, since it can be used to establish a rank-2 β tensor with only six unknowns, requiring at least six distinct diffusion encoding directions. This type of approach can reduce the amount of DW-MRI data to be acquired, and potentially serve as a viable way forward for the combined estimation of mean, axial and radial kurtosis.
+
+
+Kurtosis estimation outside of the brain
+
Although our study has been focusing on mean kurtosis imaging in the human brain, it is clear that DKI has wide application outside of the brain (Li et al., 2022b; Liu et al., 2021; Huang et al., 2021a; Guo et al., 2022a; Li et al., 2022a,d; Zhou et al., 2022). Without having conducted experiments elsewhere, we cannot provide specific guidelines for mean kurtosis imaging in the breast, kidney, liver, and other human body regions. We can, however, point the reader in a specific direction.
+
The classical mono-exponential model can be recovered from the sub-diffusion equation by setting β = 1. For this case, the product between the b-value and fitted diffusivity has been reported to be insightful for b-value sampling (Yablonskiy and Sukstanskii, 2010), in accordance with a theoretical perspective (Istratov and Vyvenko, 1999). It was suggested the product should approximately span the (0, 1) range. Considering our case based on the sub-diffusion equation, we can investigate the size of bDSUB by analysing the four non-zero b-value optimal sampling regime (Δ1: 350 s/mm2 and 1500 s/mm2; Δ2: 950 s/mm2 and 4250 s/mm2 from Table 1). Considering scGM, cGM and WM brain regions alone, the rounded and dimensionless bDSUB values are (0.09, 0.38, 0.20, 0.88), (0.13, 0.55, 0.30, 1.34) and (0.05, 0.23, 0.11, 0.48), respectively, and note that in each case the first two effective sampling values are based on Δ1, and the othertwo are derived using Δ2. Interestingly, the log-linear sampling proposed in (Istratov and Vyvenko, 1999) is closely mimicked by the effective sampling regime (scGM: -2.42, -1.62, -0.97, -0.13; cGM: -2.06, -1.20, -0.60, 0.29; WM: -2.94, -2.23, -1.48, -0.73; by sorting and taking the natural logarithm). This analysis also informs on why it may be difficult to obtain a generally large R2 across the entire brain, since β and DSUB are brain region specific and the most optimal sampling strategy should be β and DSUB specific. Whilst region specific sampling may provide further gains in the R2 value, and improve ICC values for specific brain regions, such data would take a long time to acquire and require extensive post-processing and in-depth analyses.
In biological tissue, the motion of water molecules is hindered by various microstructures, and hence the diffusion can be considerably slower than unhindered, unrestricted, free diffusion of water. The continuous time random walk formalism provides a convenient mathematical framework to model this sub-diffusive behaviour using fractional calculus (Metzler and Klafter, 2000). The resulting probability density function P(x,t) of water molecules at location x (in units of mm) at time t (in units of s) satisfies the time fractional diffusion equation:
+
+
+
+
+where
+
+ is the time fractional derivative of order β (0 < β ≤ 1) in the Caputo sense, Dβ is the generalised anomalous diffusion coefficient with unit of mm2/sβ, and the parameter β characterises the distribution of waiting times between two consecutive steps in the continuous time random walk interpretation. When β = 1, the waiting times have finite mean; when 0 < β < 1, the waiting times have infinite mean, leading to sub-diffusion behaviour. The solution to the time fractional diffusion equation (1) in Fourier space is:
+
+
+
+
+where
+
+ is the single-parameter Mittag-Leffler function, Γ is the standard Gamma function and by definition E1(z) = exp(z). In the context of diffusion DW-MRI, k in (2) represents the q-space parameter q = γGδ, t represents the effective diffusion time
+
+ and p(k, t) represents the signal intensity
+
+, leading to the diffusion signal equation (Magin et al., 2020):
+
+
+
+
+Defining
+
+, the DW-MRI signal then can be expressed in terms of b-values:
+
+
+
+
+where
+
+
+
+
+has the standard unit for a diffusion coefficient, s/mm2.
+
+
+Diffusional kurtosis imaging
+
The traditional DKI approach was proposed by Jensen et al. (Jensen et al., 2005; Jensen and Helpern, 2010) to measure the extent of non-Gaussian diffusion in biological tissues using DW-MRI data:
+
+
+
+
+where S is the signal for a given diffusion weighting b (i.e., b-value), S0 is the signal when b = 0, DDKI and KDKI are the apparent diffusivity and kurtosis. A major limitation of (6) is that it was developed based on the Taylor expansion of the logarithm of the signal at b = 0, as such b-values should be chosen in the neighbourhood of b-value = 0 (Yang et al., 2022; Kiselev, 2010). Hence, to estimate diffusivity and kurtosis, Jensen and Helpern (Jensen and Helpern, 2010) suggested the use of three different b-values (such as o, 1ooo, 2ooo s/mm2) and the maximum b-value should be in the range 2ooo s/mm2 to 3ooo s/mm2 for brain studies. Subsequently, the optimal maximum b-value was found to be dependent on the tissue types and specific pathologies, which makes the experimental design optimal for a whole brain challenging (Chuhutin et al., 2017). The procedure for fitting kurtosis and diffusivity tensors is also not trivial, and a variety of fitting procedures are currently in use. We refer readers to the descriptive and comparative studies for detail on the implementation and comparison of methods (Veraart et al., 2011b,a; Chuhutin et al., 2017).
+
+
+Mean kurtosis from the sub-diffusion model
+
Yang et al. (2022) established that the traditional DKI model corresponds to the first two terms in the expansion of the sub-diffusion model:
+
+
+
+
+
+
+where diffusivity, D*, and kurtosis, K*, are computed directly via sub-diffusion parameters DSUB and β:
+
+
+
+
+
+
+
+
+where DSUB is defined in (5). Note the mean kurtosis expression in (9) was also derived by Ingo et al. (Ingo et al., 2015) using a different method. Their derivation follows the definition of kurtosis,
+
+, i.e., by computing the fourth moment
+
+ and the second moment
+
+ based on the sub-diffusion equation (1).
+
+
+
+Connectome 1.0 human brain DW-MRI data
+
The DW-MRI dataset provided by Tian et al. (2022) was used in this study. The publicly available data were collected using the Connectome 1.0 scanner for 26 healthy participants. The first seven subjects had a scan-rescan available. We evaluated qualitatively the seven datasets, and chose the six which had consistent diffusion encoding directions. Subject 2 had 60 instead of 64 diffusion encoding directions, and hence, was omitted from this study. The 2 × 2 × 2 mm3 resolution data were obtained using two diffusion times (Δ = 19, 49 ms) with a pulse duration of δ = 8 ms and G = 31, 68, 105, 142, 179, 216, 253, 290 mT/m, respectively generating b-values = 50,350, 800, 1500, 2400, 3450, 4750, 6000 s/mm2 for Δ = 19 ms, and b-values = 200, 950, 2300, 4250, 6750, 9850, 13500, 17800 s/mm2 for Δ = 49 ms, according to b-value = (γδG)2(Δ - δ/3). 32 diffusion encoding directions were uniformly distributed on a sphere for b < 2400 s/mm2 and 64 uniform directions for b ≤ 2400 s/mm2.
+
The FreeSurfer’s segmentation labels as part of the dataset were used for brain-region specific analyses. Tian et al. (2022) provided the white matter averaged group SNR (23.10 ± 2.46), computed from 50 interspersed b-value = 0 s/mm2 images for each subject. Both magnitude and the real part of the DW-MRI were provided. Based on an in-depth analysis, the use of the real part of the DW-MRI data was recommended, wherein physiological noise, by nature, follows a Gaussian distribution (Gudbjartsson and Patz, 1995).
+
+
+Simulated DW-MRI data at specific b-values
+
DW-MRI data were simulated to establish (i) the correspondence between actual versus fitted mean kurtosis using the traditional DKI and sub-diffusion models based on various choices for Δ, and (ii) to investigate the impact of SNR levels and sub-sampling of b-values on the mean kurtosis estimate. The DW-MRI signal was simulated using the sub-diffusion model (3) with random Gaussian noise added to every normalised DW-MRI signal instance:
+
+
+
+
+where N(0, σ2) is white noise with mean of zero and standard deviation of σ according to the normal distribution.
+
Two aspects influence σ in the case of real-valued DW-MRI data. These include the SNR achieved with a single diffusion encoding direction (i.e., Connectome 1.0 DW-MRI data SNR was derived using only b-value = 0 s/mm2 data), and the number of diffusion encoding directions in each b-shell across which the powder average is computed:
+
+
+
+
+where NDIR is the number of diffusion encoding directions for each b-shell and assuming it is consistent across b-shells. The σ for the Connectome 1.0 data is approximately 1/(23.10×8) = 0.0054 based on 64 diffusion encoding directions. Achieving of SNR = 5, 10 and 20 for the simulation study can therefore be accomplished by changing only the σ and keeping NDIR = 64. As such, σSNR=5 = 0.0250, σSNR=10 = 0.0125 and σSNR=20 = 0.0063.
+
Three simulation experiments were carried out at various SNR levels. The first simulation experiment was to examine the effect of the number of diffusion times on the accuracy of the parameter fitting for idealised grey and white matter cases. In (10) the choices of Dβ = 3 × 10-4 mm2/sβ, β = 0.75, and Dβ = 5 × 10-4 mm2/sβ, β = 0.85, were made for white matter and grey matter, respectively. These two distinct βs led to K* of 0.8125 and 0.4733 using (9). Diffusion times were chosen from the range Δi ∈ [δ, δ + 50], where diffusion pulse length was set to δ = 8 ms to match the Connectome 1.0 data. A minimum required separation between any two Δs was enforced, i.e., 30/(n -1), where 30 corresponds with the 30 ms difference between the Δs for the Connectome 1.0 DW-MRI data, and n is the number of distinct diffusion times simulated. We considered as many as five distinct diffusion times. Simulations were conducted by randomly selecting sets of Δs for 1000 instances, and then generating individual DW-MRI simulated signals using (10), before fitting for Dβ and β, from which K* was computed using (9). For the case of two diffusion times, suggestions on the separation between them was given based on the goodness-of-fit of the model.
+
The second simulation experiment was to investigate the effect of the number of diffusion times on the accuracy of parameter fitting using simulated data with random values of Dβ and β. The domains were restricted to Dβ ∈ [10-4, 10-3] in the unit of mm2/sβ and β ∈ [0.5, 1], corresponding to K* ∈ [0, 1.7124]. For the case of two diffusion times, suggestions on the separation between them were given based on the goodness-of-fit of the model.
+
The third simulation experiment isto study b-value sub-sampling under various SNR levels (SNR = 5, 10 and 20). We set the b-values in the simulated data the same as those used to acquire the Connectome 1.0 dataset. At each SNR level, we selected combinations of two, three and four b-values, irrespective of the difference between them and the diffusion time set to generate the b-value. Essentially, we explored the entire possible sets of b-values for the three regimes, resulting in 120, 560 and 1820 combinations, respectively. A goodness-of-fit measure for model fitting was used to make comparisons between the different b-value combinations.
+
+
+SNR reduction by downsampling diffusion encoding directions
+
We performed SNR reduction of the Connectome 1.0 DW-MRI data by downsampling of diffusion directions in each b-shell. The method of multiple subsets from multiple sets (P-D-MM) subsampling algorithm provided in DMRITool (Cheng et al., 2018) was applied to the b-vectors provided with the Connectome 1.0 DW-MRI data. Note, the b-shells contained 32 directions if b < 2400 s/mm2, and 64 directions if b ≥ 2400 s/mm2. We consider SNR = 20 to be the full dataset. The SNR = 10 data was constructed by downsampling to eight non-collinear diffusion encoding directions in each b-shell, and three were required for the SNR = 6 data. In the downsampled data, each diffusion encoding direction was coupled with the direction of opposite polarity (i.e., SNR = 10 had sixteen measurements for each b value, and SNR = 6 had six).
+
+
+Parameter estimation
+
+Standard DKI model
+
The maximum b-value used to acquire the DW-MRI for standard DKI model fitting is limited to the range 2000 s/mm2 to 3000 s/mm2 due to the quadratic form of (6). We opted to use DW-MRI data generated with b-values = 50, 350, 800, 1500, 2400 s/mm2 using Δ = 19 ms, and b-values = 200, 950, 2300 s/mm2 using Δ = 49 ms. Note, the apparent diffusion coefficient, DDKI in (6) is time dependent, as can be deduced from (5) and (8). Thereby, standard DKI fitting can only be applied to DW-MRI data generated using a single diffusion time. The model in (6) was fitted in a voxelwise manner to the powder averaged (i.e., geometric mean over diffusion encoding directions, often referred to as trace-weighted) DW-MRI data using the lsqcurvefit function in MATLAB (Mathworks, Version 2022a) using the trust-region reflective algorithm. Optimisation function specific parameters were set to TolFun = 10-4 and TolX = 10-6. Parameters were bounded to the ranges of DDKI > 0 and 0 < KDKI ≤ 3.
+
+
+Sub-diffusion model
+
For the single diffusion time case, the sub-diffusion model in (3) was fitted to the powder averaged DW-MRI data in a voxelwise manner using the same MATLAB functions as in the previous section. For each subject, spatially resolved maps of Dβ and β were generated. The fitting strategy for multiple diffusion time data is to solve:
+
+
+
+
+where Sij is the signal at the ith effective diffusion time
+
+, and the jth q-space parameter qj, SUB is the sub-diffusion model (3) at
+
+, and qj for a given set of (Dβ,β), n is the number of diffusion times, and Ji, is the number of q-values corresponding to
+
+, in data acquisition. This objective function allows incorporation of an arbitrary number of diffusion times, each having arbitrary number of q-values. Parameters were bounded to the ranges of 0 < β ≤ 1 and Dβ > 0. Model parameters were found to be insensitive to the choice of initial values. Parameters D* and K* were computed analytically using the estimated Dβ and β according to (8) and (9).
+
+
+
+Goodness-of-fit and region-based statistical analysis
+
The coefficient of determination, referred to as R2, was used to assess how well the mean kurtosis values in the simulation were able to be fitted. It is generally accepted that an R2 value above 0.5 should be achieved and a value of 1.0 is unreasonable for data with realistic noise. Negative values imply the model is a very poor fit. For human data, we computed the region specific mean and standard deviation for each subject, and reported the weighted mean and pooled standard deviation along with the coefficient of variation (CV), defined as the ratio of the standard deviation to the mean. The weights were the number of voxels in the associated regions in each subject. Following Barrick et al. (2020), the tissue contrast is computed as
+
+, where μWM and μGM are the mean parameter values in white and grey matters; and σWM and σGM are the standard deviations of parameter values. Higher TC values indicate greater tissue contrast.
+
Human data were analysed voxelwise, and also based on regions-of-interest. We considered three categories of brain regions, namely sub-cortical grey matter (scGM), cortical grey matter (cGM) and white matter (WM). The scGM region constituted the thalamus (FreeSurfer labels 10 and 49 for left and right hemisphere), caudate (11, 50), putamen (12, 51) and pallidum (13, 52). The cGM region was all regions (1000 to 2999) and separately analysed the fusiform (1007, 2007) and lingual (1013, 2013) brain regions, while the WM had white matter fibre regions from the cerebral (2, 41), cerebellum (7, 46) and corpus callosum (CC; 251 to 255) areas. The average number of voxels in each region were 3986 (scGM), 53326 (cGM), 52121 (WM), 1634 (thalamus), 831 (caudate), 1079 (putamen), 443 (pallidum), 2089 (fusiform), 1422 (lingual), 48770 (cerebral WM), 2904 (cerebellum WM) and 447 (CC). For each brain region a t-test was performed to test for differences in mean kurtosis population means.
+
+
+Scan-rescan analysis using intraclass correlation coefficient (ICC)
+
For each of the six subjects, both the first (scan) and second (rescan) scan images were registered to the first scan images of Subject 1 using inbuilt MATLAB (Mathworks, Version 2022a) functions (imregtform and imwarp). We used 3D affine registration to account for distortions and warps common in DW-MRI data. Cubic spline interpolation was applied to resample both the scan and rescan DW-MRI data for each subject onto Subject 1’s first scan data grid. The FreeSurfer labels for Subject 1’s first scan were used for brain region analysis. The ICC measure was applied to assess scan-rescan reproducibility of mean kurtosis, as described by Duval et al. (Duval et al., 2017) and Fan et al. (Fan et al., 2021). An ICC histogram and the mean and standard deviation descriptive statistics were generated for all brain regions analysed.
+
+
+
+Conclusion
+
The utility of diffusional kurtosis imaging for inferring information on tissue microstructure was described decades ago. Continued investigations in the DW-MRI field have led to studies clearly describing the importance of mean kurtosis mapping to clinical diagnosis, treatment planning and monitoring across a vast range of diseases and disorders. Our research on robust, fast, and accurate mapping of mean kurtosis using the sub-diffusion mathematical framework promises new opportunities for this field by providing a clinically useful, and routinely applicable mechanism for mapping mean kurtosis in the brain. Future studies may derive value from our suggestions and apply methods outside the brain for broader clinical utilisation.
+
+
+
+
+Data and code availability statements
+
The Connectome 1.0 human brain DW-MRI data used in this study is part of the MGH Connectome Diffusion Microstructure Dataset (CDMD)(Tian et al., 2022), which is publicly available on the figshare repository https://doi.org/10.6084/m9.figshare.c.5315474. MATLAB codes generated for simulation study, parameter fitting, and optimising b-value sampling is openly available at https://github.com/m-farquhar/SubdiffusionDKI.
+
+
+Acknowledgement
+
Qianqian Yang and Viktor Vegh acknowledge the financial support from the Australian Research Council (ARC) Discovery Project scheme (DP190101889) for funding a project on mathematical model development and MRI-based investigations into tissue microstructure in the human brain. Qianqian Yang also acknowledges the support from the ARC Discovery Early Career Research Award (DE150101842) for funding a project on new mathematical models for capturing heterogeneity in human brain tissue. Authors also thank the members of the Anomalous Relaxation and Diffusion Study (ARDS) group for many interesting discussions involving diffusion MRI.
Nociceptive sensory neurons convey pain signals to the CNS using action potentials. Loss-of-function mutations in the voltage-gated sodium channel NaV1.7 cause insensitivity to pain (presumably by reducing nociceptor excitability) but efforts to treat pain by inhibiting NaV1.7 pharmacologically have largely failed. This may reflect the variable contribution of NaV1.7 to nociceptor excitability. Contrary to claims that NaV1.7 is necessary for nociceptors to initiate action potentials, we show that nociceptors can achieve equivalent excitability using different combinations of NaV1.3, NaV1.7, and NaV1.8. Selectively blocking one of those NaV subtypes reduces nociceptor excitability only if the other two subtypes are weakly expressed. For example, excitability relies on NaV1.8 in acutely dissociated nociceptors but responsibility shifts to NaV1.7 and NaV1.3 by the fourth day in culture. A similar shift in NaV dependence occurs in vivo after inflammation, impacting ability of the NaV1.7-selective inhibitor PF-05089771 to reduce pain in behavioral tests. Flexible use of different NaV subtypes – an example of degeneracy – compromises the reliable modulation of nociceptor excitability by subtype-selective inhibitors. Identifying the dominant NaV subtype to predict drug efficacy is not trivial. Degeneracy at the cellular level must be considered when choosing drug targets at the molecular level.
+
+SIGNIFICANCE STATEMENT
+
Nociceptors can achieve equivalent excitability using different sodium channel subtypes. The analgesic efficacy of subtype-selective drugs hinges on which subtype controls excitability. This contingency likely contributes to poor clinical outcomes.
+
+
+
+
+
+
+
+
+Competing Interest Statement
SAP is on the Scientific Advisory Boards of Boston Scientific and Presidio Medical and has received grant funding from Boston Scientific.
+
+Summary of Updates:
+
This version has been revised through addition of new text to the Introduction and Discussion. Formatting has also been adjusted, which includes changes to the organization of figures. The results themselves have not changed.
+
+
+
+
+
+INTRODUCTION
+
Chronic pain affects between 11 and 40% of the population worldwide (1). Neuropathic pain, which is pain arising from damage to the somatosensory nervous system, is particularly hard to treat with only 30% of patients achieving moderate (≥30%) relief using available treatments (2, 3). New treatments are needed but a meagre 11% of analgesic drugs entering phase 1 trials are ultimately approved (4), which has triggered debate about why basic science discoveries are not yielding improved clinical outcomes (5). Suggested explanations include flaws in preclinical animal testing (6, 7) or clinical trial design (8) but biological explanations must also be considered. For example, degeneracy – the ability of a biological system to achieve equivalent function using different components (9) – complicates modulation of neuronal excitability by allowing changes in diverse ion channels to potentially subvert the therapeutic effect of a drug targeting a particular channel (10).
+
Like most neurons, nociceptive sensory neurons (nociceptors) rely on spikes to transmit information. Their excitability is thus critical for relaying information to the CNS. Nociceptor excitability is increased in many pathological pain conditions and the resultant increase in afferent input drives chronic pain (11–13). Neuronal excitability depends on the complex interplay between diverse ion channels (14–16) but some channels seem to be particularly important for pain. For instance, loss- or gain-of-function mutations in the gene SCN9A, which encodes the voltage-gated sodium channel NaV1.7, cause congenital insensitivity to pain (CIP) or painful neuropathies, respectively (17-19; for review see 20). In rodents, nociceptor-specific deletion of NaV1.7 abolishes acute and inflammatory pain (21) but not neuropathic pain (22, 23). Neuropathic pain is blocked by deleting NaV1.7 globally, including from sympathetic neurons (24, 25), though not if the deletion is induced in adulthood (26). Furthermore, loss-of-function mutations in NaV1.7 do not consistently reduce nociceptor excitability (see Discussion) and the associated insensitivity to pain involves increased opioid signaling (27, 28), consistent with naloxone’s ability to restore pain sensitivity in CIP patients (27, 29). These observations cast doubt on whether NaV1.7 mutations produce CIP by reducing nociceptor excitability, pointing instead to a less direct mechanism that may be harder to reproduce pharmacologically.
+
Notwithstanding such reservations, several NaV1.7-selective drugs have been developed (30–32) but none have yet passed phase 2 clinical trials (33–36). This has been attributed to poor target engagement (35, 37–39) yet prevention of the flare response by PF-05198007, a NaV1.7-selective inhibitor, argues that at least some NaV1.7 channels are blocked (40). But CIP patients exhibit a normal flare response (41), suggesting that their C fibers compensate for chronic loss of NaV1.7 channels. Other NaV1.7-selective inhibitors have struggled in phase 1 trials because of autonomic side effects (e.g. 42), as might be expected if those drugs block NaV1.7 channels on sympathetic neurons, which is apparently necessary to prevent/reverse neuropathic pain (see above). But CIP patients exhibit normal autonomic function (17, 41), suggesting that their sympathetic neurons also compensate for chronic loss of NaV1.7 channels. In those patients, might similar compensation occur in nociceptors and restore pain, only for that effect to be masked by enhanced opioid signaling (see above)? Descriptions of NaV1.7 as “the” threshold channel imply that it is irreplaceable for nociceptor excitability, consistent on the surface with pain insensitivity due to loss-of-function mutations in NaV1.7 but inconsistent with some past electrophysiological data (43, 44). Clarifying whether nociceptors rely on NaV1.7 is an unresolved issue important for predicting the analgesic efficacy of NaV1.7-selective inhibitors.
+
A serendipitous observation prompted us to reassess the role of NaV1.7 in nociceptor excitability and the implications for drug efficacy. Specifically, we observed that tetrodotoxin (TTX), which inhibits NaV1.7 and several other TTX-sensitive (TTX-S) sodium channels, had variable effects in nociceptors, dramatically reducing their excitability in some conditions but not in others. This variability reveals that nociceptors can achieve equivalent excitability using different sodium channel subtypes, some of which are TTX-resistant (TTX-R). We demonstrate that a NaV1.7-selective inhibitor produces analgesia only when nociceptor excitability relies on NaV1.7. Insofar as increasingly selective drugs are more likely to have their efficacy subverted by degeneracy, our results have profound yet underappreciated implications for target selection and drug development.
+
+
+RESULTS
+
+Equivalent excitability can arise from different voltage-gated sodium (Na<sub>V</sub>) channel subtypes
+
Small dorsal root ganglion (DRG) neurons (soma diameter <25 µm) tend to spike repetitively when depolarized by current injection (45). In our sample, most small neurons genetically identified as nociceptors (see Methods) spiked repetitively when tested 2-8 hours after dissociation (DIV0) or after 4-7 days in culture (DIV4-7), though the proportion of repetitively spiking neurons increased slightly over that interval (ξ2=4.51, p=0.034, chi-square test) (Fig. 1A). Strikingly, 100 nM TTX had no effect on the spiking pattern at DIV0 but converted all but one neuron to transient spiking at DIV4-7. Amongst neurons that spiked repetitively at baseline, TTX reduced the firing rate and increased rheobase only at DIV4-7 (Fig. 1B). TTX reduced spike height at DIV0 and DIV4-7, but more so at DIV4-7. There was a significant increase in capacitance and leak conductance density between DIV0 and DIV4-7, but no change in resting membrane potential (Fig. 1C). Normalizing leak conductance by capacitance (which increases over time because of neurite growth) disambiguates whether changes in input resistance reflect changes in cell size or membrane leakiness. Consistent with current clamp data, voltage clamp recordings showed that only a small fraction of sodium current is TTX-S at DIV0, whereas nearly all sodium current was blocked by TTX at DIV4-7 (Fig. 1D). Previous studies suggested that TTX-R channels play an important role in nociceptor excitability (46–48). Our initial results confirm this for DIV0 but show that their contribution diminishes after a few days in culture, with TTX-S channels becoming dominant by DIV4. Despite this reconfiguring of NaV channels, excitability was remarkably stable, consistent with previous work showing little change in excitability after axotomy despite large (but evidently counterbalanced) changes in TTX-R and TTX-S currents (43, 49). We show later that similar changes develop in vivo following inflammation with consequences for drug efficacy assessed behaviorally (see Fig. 8), meaning the NaV channel reconfiguration described above is not a trivial epiphenomenon of culturing.
+
+
+
Different Na<sub>V</sub> subtypes produce equivalent excitability at different days in vitro (DIV).
(A) Representative responses of small DRG neurons to current injection at rheobase and 3x rheobase when tested on DIV0 (blue) or DIV4-7 (red) before (dark) and after (pale) bath application of 100 nM TTX. At DIV0, TTX did not alter spiking pattern (ξ2=0.25, p=0.617, McNemar test) or significantly reduce firing rate (F1,72=1.527, p=0.24, two-way repeated measure (RM) ANOVA; n=13). At DIV4-7, TTX significantly altered spiking pattern, converting all but one neuron to transient spiking (ξ2=20.05, p<0.0001), and it significantly reduced firing rate (F1,132=43.157, p<0.001, n=23). Only neurons with repetitive spiking at baseline are included in the firing rate plot. (B) At DIV0, TTX did not affect rheobase (Z24=1.129, p=0.265, Wilcoxon rank test) but did reduce spike height (T24=3.092, p=0.005, paired t-test). At DIV4-7, TTX increased rheobase (Z28=4.681, p<0.001, Wilcoxon rank test) and dramatically reduced spike height (T28=20.333, p<0.001, paired t-test). Notably, neurons at DIV0 and DIV4-7 did not differ in their baseline rheobase (U=316, p=0.425, Mann-Whitney test) or spike height (T52=0.322, p=0.749, t-test). (C) Neurons at DIV0 and DIV4-7 differed in their total capacitance (T52=6.728, p<0.001, t-test) and leak conductance density (U=216, p=0.011, Mann-Whitney test) but not in their resting membrane potential (T52=1.668, p=0.101, t-test). (D) Sample voltage clamp recordings with command voltage stepped from -85 mV to +15 mV in 5 mV increments, before and after TTX. Sodium current was not significantly reduced by TTX at DIV0 (F1,72=3.585, p=0.107, two-way RM ANOVA; n=7 neurons) but was completely abolished by TTX at DIV4-7 (F1,108=33.526, p<0.001; n=10 neurons). *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc tests in A and D.
+
+
+
+
+Different Na<sub>V</sub> channel subtypes control nociceptor excitability at DIV0 and DIV4-7
+
Next, we sought to identify the NaV subtype responsible for repetitive spiking at each time point, starting with DIV0. Of the TTX-R NaV channels expressed by nociceptors, NaV1.8 has been implicated in repetitive spiking (47, 48). We measured sodium current in voltage clamp before and after applying the NaV1.8-selective inhibitor PF-01247324 (PF-24) (50). At DIV0, 1 µM PF-24 abolished most of the sodium current (Fig. 2A). The PF-24-sensitive current had slow inactivation kinetics, like the TTX-R current and unlike the fast TTX-S current in Figure 1D, and consistent with previous descriptions of NaV1.8 (51). A different Nav1.8 antagonist, A-803467, had similar effects (Fig. 2 – figure supplement 1). In current clamp, PF-24 converted 7 of 8 repetitively spiking neurons to transient spiking and significantly reduced evoked spiking (Fig. 2B). It also increased rheobase and decreased spike height but did not affect resting membrane potential (Fig. 2C). PF-24 had negligible effects when tested at DIV4-7 (Fig. 2 – figure supplement 2). These results show that NaV1.8 is the predominant NaV subtype at DIV0 and is necessary for repetitive spiking at that time point. To test the sufficiency of NaV1.8 to produce repetitive spiking, we tuned a conductance-based model neuron (see Methods) to reproduce DIV0 data described above. In this DIV0 model, inclusion of NaV1.8 conductance was sufficient to generate repetitive spiking (Fig. 2D left). The necessity of NaV1.8 for repetitive spiking at DIV0 was also recapitulated: 85% reduction in the NaV1.8 conductance converted spiking from repetitive to transient (Fig. 2D and Supplementary Table 1).
+
+
+
Na<sub>V</sub>1.8 is necessary for repetitive spiking at DIV0.
(A) Sample voltage clamp recordings show that sodium current was almost completely abolished by the NaV1.8 inhibitor PF-24 (1 µM). Peak current was significantly reduced by PF-24 (F1,72=12.651, p<0.012, two-way RM ANOVA; n=7). Another NaV1.8 inhibitor, A-803467, had a similar effect (see Fig. 2 – figure supplement 1). (B) PF-24 significantly altered spiking pattern (χ2=5.14, p=0.0233, McNemar test) and reduced firing rate (F1,42=11.946, p=0.011, two-way RM ANOVA; n=8). (C) PF-24 significantly increased rheobase (Z15=2.783, p=0.003, Wilcoxon rank test) and reduced spike height (T15=3.151, p=0.007, paired t-test) but did not affect resting membrane potential (T15=0.304, p=0.765, paired t-test). PF-24 had limited effects at DIV4-7 (Fig 2 – figure supplement 2). (D) A computational model reproduced the effect of NaV1.8 on spiking pattern (also see Supplementary Table 1). The PF-24 effect was simulated as a ∼85% reduction in NaV1.8 (𝑔̄Nav1.8== 4mS/cm2). *, p<0.05; **; p<0.01; Student-Newman-Keuls post-hoc tests in A and B.
+
+
+
Next, we sought to identify the NaV subtype responsible for repetitive spiking at DIV4-7 using PF-05089771 (PF-71) to inhibit NaV1.7 (40, 52) and ICA-121431 (ICA) to inhibit NaV1.1/1.3 (53, 54). In voltage clamp, sodium current was significantly reduced by 30 nM PF-71, and most of the remaining current was blocked by 1 µM ICA (Fig. 3A). In current clamp, each inhibitor (applied separately) converted a significant proportion of neurons to transient spiking and significantly reduced firing rate (Fig. 3B). This argues that NaV1.7 and NaV1.1/1.3 are both necessary for repetitive spiking at DIV4-7. Inhibiting NaV1.7 increased rheobase, unlike inhibiting NaV1.1/1.3, and caused a stronger reduction in spike height (Fig. 3C). Neither affected resting membrane potential. These results show that NaV1.7 is the predominant NaV subtype at DIV4-7, but not the only one. PF-71 had negligible effects when tested at DIV0 (Fig. 3 – figure supplement 1). We re-tuned our computational model to reproduce DIV4-7 data, with both NaV1.7 and NaV1.3 being required to produce repetitive spiking, meaning neither channel is individually sufficient (Fig. 3D and Supplementary Table 1). That said, inserting a higher density of either NaV1.7 or NaV1.3 could produce repetitive spiking in the absence of the other subtype (Fig. 3 – figure supplement 2), consistent with NaV1.7 and NaV1.3 also being interchangeable.
+
+
+
Na<sub>V</sub>1.3 and Na<sub>V</sub>1.7 are necessary for repetitive spiking at DIV4-7.
(A) Sample voltage clamp recordings show that sodium current was reduced by the NaV1.7 inhibitor PF-71 (30 nM) and by the NaV1.1/1.3 inhibitor ICA (1 µM). Peak current was significantly reduced by PF-71 and ICA (F2,192=26.361, p<0.001, two-way RM ANOVA; n=9). (B) PF-71 and ICA both significantly altered spiking pattern (χ2=4.17, p=0.041 and χ2 =7.11, p=0.0077, respectively, McNemar tests) and significantly reduced firing rate (F1,54=40.659, p<0.001, n=10 and F1,78=35.156, p<0.001, n=14, respectively, two-way RM ANOVAs). (C) PF-71 significantly increased rheobase (Z18=3.464, p<0.001, Wilcoxon rank test) and decreased spike height (T18=7.946, p<0.001, paired t-test). ICA did not significantly alter rheobase (Z18=1.248, p=0.225) but did reduce spike height (T18=3.243, p=0.005). Neither drug affected resting membrane potential (T15=1.681, p=0.113 for PF-71; T18=-1.132, p=0.272 for ICA, paired t-test). PF-71 had negligible effects at DIV0 (Fig. 3 – figure supplement 1). (D) A computational model reproduced the combined effects of NaV1.3 and NaV1.7 on spiking pattern (also see Supplementary Table 1 and Fig. 3 – figure supplement 2). PF-71 effect was simulated as a 70% reduction in NaV1.7 (𝑔̄Nav1.7= 10.5 mS/cm2). ICA effect was simulated as a 90% reduction in NaV1.3 ((𝑔̄Nav1.3= 0.035 mS/cm2). *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc tests in A and B.
+
+
+
+
+Acutely interchanging Na<sub>V</sub> subtypes does not affect spiking pattern
+
The ability of NaV1.3, NaV1.7 and NaV1.8 to each encourage repetitive spiking is seemingly inconsistent with the common view that each NaV subtype contributes selectively to a different phase of the spike (for example, Figure 3 in ref 55). If NaV1.8 were to activate exclusively at suprathreshold voltages, it could not initiate spikes and a different perithreshold-activating NaV channel would be needed, which is clearly inconsistent with our data. To verify that NaV1.7 and NaV1.8 currents are each sufficient to produce repetitive spiking, we tested whether the NaV1.8 current necessary for spiking in our DIV0 model could be replaced with NaV1.7, and whether the NaV1.7 current necessary for spiking in our DIV4-7 model could be replaced with NaV1.8. In both cases, repetitive spiking was restored after inserting the alternate current (Fig. 4A). We then proceeded with equivalent experiments in real neurons, inhibiting NaV1.8 with PF-24 on DIV0 or NaV1.7 with PF-71 on DIV4-7, and then introducing the alternate channel virtually using dynamic clamp (see Methods). The replacement was successful in all neurons tested (Fig. 4B). Inserting virtual NaV1.8 after inhibiting native NaV1.8 also restored repetitive spiking, and likewise for NaV1.7 (Fig. 4 – figure supplement 1), verifying that our virtual channels were equivalent to the native channels we aimed to replace. Apart from maximal conductance density, which was titrated in each neuron, all other parameters used for dynamic clamp were identical to simulations. The success of dynamic clamp experiments helps validate our computational models insofar as virtual NaV1.7 and NaV1.8 currents interacted appropriately with native currents to produce repetitive spiking in real neurons, the same way they interact with other simulated currents in the model neuron.
+
+
+
Na<sub>V</sub>1.7 and Na<sub>V</sub>1.8 are each sufficient to produce repetitive spiking in DIV0 and DIV4-7 neurons.
(A) The computational model predicts that the NaV1.8 conductance, which is “necessary” for repetitive spiking at DIV0 can, in principle, be replaced by NaV1.7 (left), and vice versa at DIV4-7 (right). (B) Replacement experiments involved inhibiting native channels pharmacologically and then introducing virtual conductances using dynamic clamp. At DIV0 (left), inhibiting native NaV1.8 (with PF-24) converted neurons to transient spiking, but introducing virtual NaV1.7 reverted neurons to repetitive spiking (in 3 of 3 neurons tested). At DIV4-7, inhibiting native NaV1.7 (with PF-71) converted the neuron to transient spiking, but introducing virtual NaV1.8 reverted neurons to repetitive spiking (in 4 of 4 neurons tested). Repetitive spiking was likewise restored by replacing the blocked native channel with the corresponding virtual channel (Fig. 4 – figure supplement 1). Parameters for virtual channels were identical to simulations except for the maximal conductance density, which was titrated in each cell.
+
+
+
With the model neurons thus validated, we used simulations to measure NaV1.7 and NaV1.8 currents during different phases of the spike (Fig. 5A-D). Since inward (depolarizing) current at voltages near spike threshold is critical for spike initiation (56; see above), we sought to identify which NaV contributes to that current. In the DIV0 model, NaV1.7 dominated during onset of the first spike but all subsequent spikes were initiated by NaV1.8 (Fig. 5A,B). This occurred because the small NaV1.7 conductance at DIV0 quickly inactivated during the first spike and remained inactive during subsequent spikes (Fig. 5 – figure supplement 1A). This is consistent with experimental results, where repetitive spiking at DIV0 was unaffected by inhibiting NaV1.7 (see Fig. 1 and Fig. 3 – figure supplement 1) but was prevented by inhibiting NaV1.8 (see Fig. 2). However, inactivation of NaV1.7 after the first spike was reflected by an increase in voltage threshold between the first and second spike, both in the model (Fig. 5A) and in experiments (Fig. 5E). This unexpected simulation result also predicted that TTX should affect the voltage threshold of the first spike in DIV0 neurons despite not having other notable effects (see Fig. 1); as predicted, TTX caused a significant depolarizing shift in voltage threshold at DIV0 (Fig. 5 – figure supplement 2), further validating our model. In the DIV4-7 model, NaV1.7 and NaV1.3 contributed to initiation of all spikes whereas NaV1.8 was negligible (Fig. 5C,D). Even though inactivation reduced NaV1.7 and NaV1.3 current after the first spike (Fig. 5 – figure supplement 1B), those channels nonetheless provided sufficient inward current to support repetitive spiking at DIV4-7. Inactivation at DIV4-7 was reflected, however, in a combination of higher threshold and lower spike overshoot for the second spike, both in the model (Fig. 5C,D) and in experiments (Fig. 5E). These results demonstrate that each NaV subtype does not contribute exclusively to a particular phase of the spike, and nor is each spike phase mediated exclusively by a particular NaV subtype; instead, each subtype contributes preferentially to a different spike phase depending on its voltage-dependency, but conductance density and inactivation status are both important. Indeed, an subtype’s contribution can shift rapidly (because of channel inactivation) or slowly (because of gene expression changes).
+
+
+
Contribution of Na<sub>V</sub>1.7 and Na<sub>V</sub>1.8 to spike initiation in DIV0 and DIV4-7 neurons.
(A) Voltage (top) for first (left) and second (right) spikes in the DIV0 model aligned with voltage activation curves for each NaV subtype (bottom). Dashed line shows voltage threshold (defined as V where dV/dt reaches 5 mV/ms). (B) Conductance plotted against voltage to create a phase portrait (top) showing NaV conductance at different phases of the spike. Inset shows full voltage range; main graph zooms in on voltages near threshold. Bottom plots show current plotted over the same voltage range. Whereas NaV1.7 (orange) mediated nearly all perithreshold inward current for the first spike, voltage threshold increased – because NaV1.7 inactivated (Fig. 5 – figure supplement 1) – and NaV1.8 (green) mediated nearly all perithreshold inward current for the second spike. The unexpected contribution of NaV1.7 to the first spike correctly predicted that TTX increases voltage threshold in DIV0 neurons (Fig. 5 – figure supplement 2). (C, D) In the DIV4-7 model, NaV1.7 (orange) and NaV1.3 (maroon) contributed to initiation of all spikes whereas the contribution of NaV1.8 was negligible (due entirely to its low expression level). (E) Sample experimental traces showing differences in the first (blue/red) and second (grey) spikes at DIV0 and DIV4-7. Plots summarize differences (ý) in threshold, overshoot potential, and spike rise time between 1st and 2nd spikes during repetitive spiking evoked by current injection. At DIV0, the 1st and 2nd spikes differ significantly in their threshold (T8=2.522, p=0.036, one-sample t-test) and overshoot (T8=0.038, p=0.038) but not rise time (T8=0.249, p=0.810). At DIV4-7, the 1st and 2nd spikes differ in all measures (threshold: T7=7.613, p<0.001; overshoot: T7=-9.849, p<0.001; rise time: T7=5.979, p<0.001). Statistical results (green) show that differences between 1st and 2nd spike at DIV4-7 are significantly larger than differences at DIV1 (threshold: T15=-3.847, p=0.002; overshoot: T15=7.922, p<0.001; rise time: T15=-5.617, p<0.001, unpaired t-tests), consistent with our computational model.
+
+
+
+
+Control of changes in Na<sub>V</sub> subtype expression between DIV0 and DIV4-7
+
Next, we sought to identify the basis for the slow shift in which NaV subtype controls nociceptor excitability. Figure 6A shows mRNA levels for NaV1.7 and NaV1.8 relative to a housekeeping gene (left) and to each other (right). NaV1.7 mRNA levels exceeded NaV1.8 mRNA levels at both at both DIV0 and DIV7. Both decreased between DIV0 and DIV7, but NaV1.8 more so, resulting in a significant decrease in the NaV1.8:NaV1.7 mRNA ratio. This pattern is consistent with the reduced role of NaV1.8 at DIV4-7 but is inconsistent with the negligible role of NaV1.7 at DIV0; specifically, we expected NaV1.7 mRNA levels to increase between DIV0 and DIV7. Next, we investigated if functional changes were better reflected by changes in protein levels. Immunofluorescence for NaV1.8 was higher than for NaV1.7 at DIV0, and that ratio reversed at DIV7 (Fig. 6B), consistent with functional changes. Moreover, cercosporamide (10 µM), a potent inhibitor of the eukaryotic translation Initiation Factor 4E (eIF4E), significantly mitigated the decrease in NaV1.8 immunofluorescence and the increase in NaV1.7 immunofluorescence when applied to cultured neurons for 24 or 120 hours prior to measurements on DIV5 (Fig. 6C). Beyond showing that their mRNA levels do not correlate well with NaV contributions to nociceptor excitability, reminiscent of some previous work (e.g. 57), these results suggest that translational regulation is crucial, though membrane trafficking and other downstream processes likely also contribute (58, 59).
+
+
+
Protein levels, but not mRNA, reflect functional contributions of Na<sub>V</sub> subtypes at DIV0 and DIV7.
(A) Both NaV1.8 and NaV1.7 mRNA levels (relative to a housekeeping gene (HKG), see Methods) decreased significantly between DIV0 and DIV4-7 (factor 1: time, F1,12=56.677, p<0.001, factor 2: subtype, F1,12=17.952, p=0.001, two-way ANOVA and Student-Newman-Keuls post-hoc tests on log transformed data, n=4 mice per time point) but more so for NaV1.8 than for NaV1.7 (interaction: time x subtype, F1,12= 11.455, p=0.005). The differential reduction yielded a significantly higher NaV1.8: NaV1.7 ratio at DIV0 than at DIV7 (T6=21.375, p<0.001, unpaired t-test) but the increasing functional contribution of NaV1.7 between DIV0 and DIV4-7 remains unaccounted for. (B) Immunoreactivity (IR) for NaV1.8 protein exceeded NaV1.7-IR at DIV0, but the opposite was true on DIV4-7, consistent with the functional contribution of each subtype. NaV-IR was measured relative to YFP intensity in the same cell, and then each cell’s NaV1.8:YFP ratio was considered relative to the average NaV1.7:YFP ratio in the co-processed coverslip (left) or average NaV1.8:YFP ratio was considered relative to the average NaV1.7:YFP ratio in the same animal (right). Ratios were >1 at DIV0 but decreased significantly at DIV4-7 (U=78, p<0.001, n=37 for DIV0, n=40 for DIV4-7, Mann-Whitney test (left) and T6=4.046, p=0.007, unpaired t-test (right)). (C) Chronically applied cercosporamide (10 µM) mitigated the changes in NaV1.8-and NaV1.7-IR at DIV5 (NaV1.8: H3=157.95, p<0.001; NaV1.7: H3=80.662, p<0.001; One-way ANOVA on ranks, Dunn’s post-hoc tests, p<0.05 for all pairs). Panel on the right shows data normalized to baseline (DIV0) to emphasize relative changes.
+
+
+
+
+Analgesic efficacy of subtype-selective drugs depends on which Na<sub>V</sub> controls nociceptor excitability
+
If a NaV1.7-selective inhibitor mediates analgesia by modulating nociceptor excitability, its analgesic efficacy hinges on nociceptor excitability being controlled by NaV1.7. Accordingly, we predicted that the NaV1.7-selective inhibitor PF-71 would have little if any effect on paw withdrawal under normal conditions, when NaV1.8 controls nociceptor excitability (Fig. 2 and Fig. 3 – figure supplement 1), but would be effective if NaV1.7 took over control. Inflammation increases NaV1.7 channel trafficking and membrane expression (60–63). To test if inflammation increased NaV1.7’s influence on nociceptor excitability, we recorded neurons acutely dissociated (DIV0) from DRGs of mice whose hind paw was injected with CFA three days prior. Inflammation caused nociceptors to become much more variable in their reliance of specific NaV subtypes (Fig. 7A). Despite this variability, inhibiting NaV1.7 with PF-71 converted a significantly higher proportion of neurons to transient spiking after CFA (42%) than in control neurons (0%) (Fig. 7B, left); subsequent application of PF-24 to inhibit NaV1.8 converted just 14% of CFA neurons to transient spiking vs 88% of control neurons (Fig 7B, right). PF-71 also significantly affected resting membrane potential, rheobase, and spike height after CFA (Fig. 7C), unlike in control neurons (see Fig. 3 – figure supplement 1).
+
+
+
Inflammation alters Na<sub>V</sub> subtype contribution to nociceptor excitability.
(A) Sample responses in DIV0 neurons from mice injected with CFA three days earlier. In 12 cells tested, PF-71 converted 5 neurons to transient spiking (i), encouraged repetitive spiking in 4 neurons (ii), and had no effect in 3 neurons (iii), thus highlighting increased heterogeneity after CFA. (B) At DIV0, the effect of PF-71 differed significantly between CFA and control neurons, converting 42% (5 of 12) CFA neurons from repetitive to transient spiking vs 0% (0 of 9) control neurons (p=0.0451, Fisher Exact test). Applying PF-24 to neurons that continued to spike repetitively after PF-71 had little effect on CFA neuron, converting only 13% (1 of 7) of CFA neurons vs 88% (7 of 8) of control neurons (p=0.001, Fisher Exact test). Together these results argue that NaV1.7 contributes more and NaV1.8 contributes less to nociceptor excitability after inflammation. (C) At DIV0, PF-71 significantly increased resting membrane potential (T11=-3.530, p=0.005, paired t-test) and rheobase (Z11=2.186, p=0.024, Wilcoxon rank test), and significantly decreased spike height (T11=4.413, p=0.001, paired t-test) in CFA neurons. Further addition of PF-24 significantly changed rheobase (Z9=2.176, p=0.023, Wilcoxon rank test) and spike height (T9=3.237, p=0.01, paired t-test) but did not affect resting membrane potential (T9=1.049, p=0.321, paired t-test).
+
+
+
Results above confirm that NaV1.7 takes on greater responsibility for nociceptor excitability after inflammation, which in turn predicts that PF-71 should reduce pain after inflammation but not under control conditions. As predicted, PF-71 significantly reduced thermal (Fig. 8A) and tactile (Fig. 8B) sensitivity in CFA-inflamed mice without having any effect in control mice. Consistent with this, epigenetic repression of NaV1.7 prevents/reverses hypersensitivity in inflamed and neuropathic mice without causing hyposensitivity in naïve mice (64). This is unlike genetic deletion of NaV1.7, which reduces thermal and tactile sensitivity in naïve mice (27), and with loss-of-function mutations in NaV1.7 that abolish pain in humans (17). These inconsistencies rekindle concerns whether NaV1.7 mutations, unlike pharmacological interventions, affect pain through mechanisms other than modulation of nociceptor excitability. Pharmacological reversal of hypersensitivity in chronic pain conditions (when NaV1.7 is pathologically upregulated) without reducing normal nociceptive pain is clinically desirable, but this hinges on nociceptor hyperexcitability being NaV1.7-dependent, which may be true of some but not all chronic pain conditions, or in only a subset of patients (65).
+
+
+
+DISCUSSION
+
Our results show that nociceptors can achieve similar excitability using different NaV channels. Whereas repetitive spiking depends on NaV1.8 shortly after dissociation (Fig. 2) and presumably under normal conditions in vivo, responsibility shifts to NaV1.7 and NaV1.3 after a few days in vitro (Fig. 3). This is due to translationally regulated changes in NaV expression (Fig. 6). Inflammation causes a similar shift in vivo (Fig. 7). Importantly, acutely inhibiting a particular NaV is consequential (analgesic) only if that subtype is responsible for nociceptor excitability (Fig. 8). This may explain why NaV1.7-selective drugs have not performed well in clinical trials (see Introduction) – because NaV1.7 is not always necessary for nociceptor excitability depending on the expression level of NaV1.7 and other NaV subtypes. Faster processes like channel inactivation also affect their relative contribution (Fig. 5). These observations demonstrate the variable contribution of different NaV subtypes to nociceptor excitability. When unaccounted for, such variability can lead to inconsistencies at the root of poor reproducibility and translatability.
+
+
+
Inflammation-induced change in Na<sub>V</sub> subtype contribution impacts analgesic efficacy of PF-71.
(A) CFA significantly increased thermal sensitivity (F5,65=19.556, p<0.001, two-way RM ANOVA). PF-71 significantly decreased thermal sensitivity in mice injected three days prior with CFA (T8=-7.296, p<0.001; paired t-test) but had no effect in naïve mice (T5=-0.141, p=0.894). (B) CFA significantly increased mechanical sensitivity (F4,52=16.786, p<0.001). PF-71 significantly decreased tactile sensitivity in mice injected three days prior with CFA (T8=-4.341, p=0.002) but had no effect in naive mice (T5=1.000, p=0.363). Insets in both panels show values for each animal before and 2 hours after PF-71 injection. *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc tests.
+
+
+
Contrary to the view that certain ion channels are uniquely responsible for certain aspects of neuronal function, neurons use diverse ion channel combinations to achieve similar function (66, 67). This degeneracy is crucial for enabling excitability and other aspects of neuron physiology to be homeostatically regulated by adjusting ion channels in response to perturbations (68–70). Degeneracy also enables pathological changes in different ion channels to produce equivalent hyperexcitability (71). This is important insofar as similar excitability may belie differences in the underlying ion channels – differences that may render a neuron susceptible or impervious to a drug depending on the functional necessity of the targeted ion channel in that neuron. This is precisely what our data demonstrate in nociceptors. Similar observations have been made in substantia nigra neurons, whose pacemaker activity can be mediated by NaV channels or by voltage-gated calcium channels, meaning TTX may or may not block their spiking (71, 72). Similar interchangeability is evident for the burst firing of Purkinje neurons (73).
+
Degeneracy also exists at the circuit level (74, 75), where it allows differences in the intrinsic excitability of component neurons to be offset (and effectively hidden) by differences in synaptic weights (76). Relevant for pain processing, the spinal dorsal horn circuit can achieve similar output using different synaptic weight combinations (77); specific neuron types may have a greater or lesser impact on circuit function depending on those weights. In effect, degeneracy introduces contingencies. The role of any ion channel in a neuron (or any neuron in a circuit) depends on the other ion channels in that neuron (or the synaptic connections with other neurons in the circuit). Because of such contingencies, a drug may engage its target without producing the intended cellular, circuit or clinical effect. Indeed, different combinations of GABAA receptor activation and chloride driving force can produce equivalent synaptic inhibition (78), but when inhibition is incompensably compromised, the underlying cause necessitates different interventions (79). By this logic, if upregulation of NaV1.7 is responsible for nociceptor hyperexcitability after nerve injury or inflammation, NaV1.7 is an ideal target since “normal” neurons (not reliant on NaV1.7) would be spared the effects of a NaV1.7-selective drug, but the long-term efficacy of such a drug hinges on hyperexcitability remaining NaV1.7-dependent, which cannot be assumed (10). Furthermore, if myelinated afferents (which express minimal NaV1.7 (80)) are responsible for mechanical allodynia under neuropathic conditions (81–84), then NaV1.7-selective drugs should not be expected to alleviate that symptom, which evidently they do not (26), at least not through a direct mechanism. Indeed, ablating nociceptors abolishes acute and inflammatory pain but not neuropathic pain (23, 85). Pathological pain being mediated by more than one afferent type is another example of circuit-level degeneracy.
+
To be interchangeable, NaV subtypes must functionally overlap (86, 87). Indeed, NaV1.8 and NaV1.7 are similar but not identical in their gating properties; for example, their voltage-dependencies partially overlap but the activation curve for NaV1.8 is right-shifted compared to NaV1.7 (88). Consequently, NaV1.7 activates at voltages near threshold whereas NaV1.8 tends to activate at suprathreshold voltages, during initiation and upstroke of the spike, respectively (34, 55). But that separation is not absolute. We found that NaV1.7 contributes to initiation of the first spike in DIV0 neurons, but because it inactivates more readily than NaV1.8, initiation of all subsequent spikes depends on NaV1.8 (see Fig. 5), which activates at perithreshold voltages because voltage threshold is high (depolarized) in the absence of NaV1.7. At DIV4-7, NaV1.7 still inactivates (which causes voltage threshold to rise) but, because of its higher density, continues to produce enough inward current to continue to initiate later spikes. The activation pattern we report for the first spike at DIV0 is consistent with Blair and Bean (89) who quantified the contribution of different NaV channels by recording pharmacologically isolated currents while varying the holding potential according to the spike waveform. Our results go further in showing how responsibilities shift across different spikes within a train (because of differential NaV inactivation) and across conditions (because of changes in NaV expression).
+
With respect to reproducibility, labs testing nociceptors after different times in vitro would be expected to reach contradictory conclusions about the relative importance of a given NaV subtype. Likewise, a testing protocol focusing on single spikes (the equivalent to the first spike in a train) would yield different results from one that considers repetitive spiking. Along the same lines, voltage clamp protocols that deliberately hold membrane potential at unnaturally hyperpolarized voltages to relieve inactivation before stepping up the voltage can give a misleading impression of how much a NaV subtypes contributes under natural conditions (i.e. with natural levels of inactivation). Such discrepancies might be chalked up to irreproducibility if the consequences of those methodological differences are not appreciated, especially if one overlooks how degeneracy allows responsibilities to shift between ion channels. Indeed, the pain literature is replete with apparent inconsistencies. We would argue that most of those studies are correct, but only under limited conditions; failure to identify and report those conditions (contingencies) represents a huge impediment to translation.
+
In summary, our results show that nociceptors can achieve equivalent excitability using different NaV subtypes. The importance of a given subtype can shift on long and short timescales, yielding results that are seemingly inconsistent. By elucidating those shifting responsibilities, our results highlight the degenerate nature of nociceptor excitability and its functional implications. Degeneracy makes it impossible to claim without reservation that a particular NaV subtype is uniquely responsible for pathological pain. Greater appreciation of degeneracy’s implications would prompt better experimental design, more cautious interpretation, and, ultimately, improved translation.
+
+
+MATERIALS AND METHODS
+
+Animals
+
All animal procedures were approved by the Animal Care Committee at The Hospital for Sick Children (protocol #53451) and were conducted in accordance with guidelines from the Canadian Council on Animal Care. We used the Cre-loxP recombinase system to generate mice that express ChR2-eYFP in NaV1.8-expressing neurons. Mice were obtained by crossing homozygous Ai32 mice (B6.Cg-Gt(ROSA)26Sortm32(CAG-COP4*H134R/EYFP)Hze/J) from Jax (#012569), which express ChR2(H134R)-eYFP in the presence of Cre recombinase, with NaV1.8-Cre mice (Tg(Scn10a-cre)1Rkun), which express Cre recombinase in Nav1.8-expressing neurons (kindly provided by Rohini Kuner). These neurons are primarily nociceptive and thermoreceptive (90). The NaV1.8 promoter leads to transgene expression in >90% of neurons expressing markers of nociceptors (21). To ensure that our transgenic mice were typical of wild-type mice with the same background (C57BL/6j), experiments reported in Fig. 1 were repeated in both genotypes for comparison. There was no effect of genotype on rheobase, spike height, input resistance, or spiking pattern, nor was there any significant interaction between genotype and effects of TTX except for spike height at DIV4-7, where TTX had a marginally larger effect in wild-type mice (Two-way ANOVA, F1,54=4.968, p=0.03, see source data file); therefore, we pooled the data for Fig.1. Having verified that our foundational observations held across different genotypes, we used transgenic mice for all subsequent experiments in order to identify eYFP-expressing nociceptors for patching, collection, or imaging.
+
+
+Dorsal root ganglia neuron cultures
+
All key reagents are listed in Supplementary Table 2. Methods for primary DRG culture have been described previously (91). Briefly, adult mice (> 7 week-old) were anaesthetised with isoflurane and perfused intracardiacally with cold HBBS (without Ca and Mg, LifeTech 14170112) supplemented with (in mM) 15 HEPES, 28 Glucose, 111 sucrose, and pH adjusted with NaOH to 7.3-7.4; osmolarity 319-321. Lumbar dorsal root ganglia (DRGs) were extracted (L2-5, except for CFA-inflamed mice, in which we only took L4), digested with papain (Worthington Biochemical Corp.) and collagenase (Worthington Biochemical Corp.)/dispase II (Sigma), and mechanically dissociated by trituration before being plated onto poly-D lysine-coated coverslips and incubated in Neurobasal™ media (Gibco 21103-049) supplemented with 1 % fetal bovine serum (FBS), B-27™ supplement (Thermo Fisher 17504-044) and 0.5mM L-Glutamine (Gibco 25030-081) for an initial period of 2 hours. After this, media was changed to maintenance media (same as plating media but without FBS) and cells were maintained in a 5% CO2 incubator at 37°C. Media was changed every 3-4 days thereafter. Neurons were recorded at two time points after plating: 2-8 hours (referred to as DIV0) or 4-7 days (referred to as DIV4-7).
+
+
+Electrophysiology
+
Coverslips with cultured neurons were transferred from the incubator to a recording chamber perfused with artificial cerebrospinal fluid containing (in mM): 126 NaCl, 2.5 KCl, 2.0 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, 2 MgCl2, and 10 glucose, bubbled with carbogen (5% CO2:95% O2) at room temperature. Neurons were visualized with gradient contrast optics on a Zeiss AxioExaminer microscope using a 40x, 0.75 NA water immersion objective (N-Achroplan, Zeiss) and IR-1000 Infrared CCD camera (Dage-MTI). YFP expression was visualized by epifluorescence (X-Cite, Excelitas) using a Zeiss filter set (46HE). A long-pass filter (OG590) was positioned in the transmitted light path to avoid activating ChR2 while patching. No optogenetic testing was performed as part of this study. Cells expressing YFP and with a soma diameter <25 µm were targeted for whole cell recording using pipettes (∼5 MΩ resistance) pulled from borosilicate glass (WPI).
+
For current clamp recordings, pipettes were filled with intracellular solution containing (in Mm): 140 K-gluconate, 2 MgCl2, 10 HEPES, 0.2 EGTA, 3.8 Na-ATP and 0.4 Na-GTP with pH adjusted to 7.3 with KOH; osmolarity was ∼300 mOsm. A liquid junction potential correction of 15 mV was applied to all reported voltages. Series resistance was compensated to >70%. Signals were amplified with an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, USA), low-pass filtered at 2 kHz, digitized with a Power1401 A/D device (Cambridge Electric Design, Cambridge, UK), and recorded at 10 kHz using CED software Signal version 6. After the natural resting membrane potential was noted, neurons were adjusted to -70 mV using continuous current injection in current clamp mode. Action potentials (spikes) were evoked using a series of 1-second long depolarizing current injections. Rheobase was defined as the minimal current required to evoke a spike. Neurons were tested with current injections from 1x rheobase to 4x rheobase using increments of 0.5x rheobase. Repetitive spiking neurons were defined as those producing ý3 spikes in response to any stimulus intensity; transient spiking neurons consistently produced ≤2 spikes. Spike threshold was defined as voltage where dV/dt first exceeds 5 mV/ms (92). Spike height was measured from threshold to peak of the action potential. Only neurons with a resting membrane potential below -45 mV, spikes overshooting 0 mV and recordings with <20% change in series resistance were tested and analyzed. For dynamic clamp experiments, the pipette shank was painted with Sylgard (Dow) to reduce pipette capacitance. Virtual NaV1.7 and NaV1.8 conductances were introduced into the cells using CED software Signal v6. Currents were defined using the Hodgkin-Huxley equation, using the same parameter values as in our computational model (see below).
+
For voltage-clamp recordings, the bath solution was adjusted to reduce sodium currents to ensure proper clamping (85). Bath solution contained (in mM): 65 NaCl, 50 choline chloride, 5 KCl, 5 HEPES, 5 MgCl2, 10 glucose, and 0.1 CaCl2, plus 0.1 CdCl2 to block calcium currents, and 20 TEA and 5 4-AP to block potassium currents; pH was adjusted to 7.4 with NaOH. Pipettes where filled with intracellular solution containing (in Mm): 140 CsCl, 10 HEPES, 2 MgCl2, 1 EGTA, 3.8 Na-ATP, 0.4 Na-GTP; pH was adjusted to 7.3 with CsOH. The resulting pipette resistance was ∼3 MΩ. A liquid junction potential correction of 4.8 mV was applied to all command voltages. Sodium currents were recorded during 20 ms-long steps from -85 mV to voltages between -45 and +15 mV. Series resistance was compensated to >80%. Signals were amplified, low-pass filtered at 5 kHz, and digitized as described for current clamp recordings.
Cultured DRG neurons <25 μm and expressing eYFP were identified as described above for patching. Coverslips were perfused with aCSF made with DEPC-treated ddH2O, and identified neurons were collected using a glass pipette filled with intracellular solution also made from DEPC-treated ddH2O (composition otherwise the same as described above for electrophysiology). Approximately 50 neurons were collected at DIV0 and at DIV4-7. Total mRNA was extracted with a PureLink RNA mini kit after digestion of genomic DNA with DNase I (Thermo Fisher Scientific) and the cDNA was synthesized with a SuperScript II first-strand synthesis kit (Thermo Fisher Scientific) according to instructions. RT-qPCR was performed with the cDNA primers of target genes (Supplementary Table 3), and the PowerUp SYBR® Green master mix (Thermo Fisher Scientific) in the QuantStudio-3 real-time PCR system. The primers were designed with IDT and spanned at least one exon longer than 1000 bp in order to exclude contamination from genomic RNA. Non-RT mRNA was also used as a negative control to exclude contamination from genomic RNA. All target genes were performed in triplicate for each sample and the experiments were repeated at least 3 times. NaV1.7 and NaV1.8 transcript levels were analyzed using the 2-ΔΔCT method and compared with the housekeeping gene HPRT.
+
+
+Immunocytochemistry
+
Cultured DRG neurons were treated with 4% paraformaldehyde for 10 minutes, rinsed 3x with cold PBS, and permeabilized with 0.1% Triton X-100 in PBS. After another 3x rinse with PBS, neurons were treated with 10% normal goat serum for 30 min followed with rabbit primary NaV1.7 antibody (1:200, ASC-008, Alomone) or NaV1.8 antibody (1:200, ASC-028, Alomone) in PBS with 0.1% Tritween-20 and 1% BSA for 1 h. For some of the coverslips, primary antibodies were replaced with control peptides (ASC008AG1040 for NaV1.7 and ASC016AG0640 for NaV1.8) provided by Alomone as negative controls. Following 3x rinse in PBS, neurons were incubated in the dark with goat anti-rabbit secondary antibody Alexa Fluor-647 (1:500, Abcam) in PBS containing 1% BSA for 1 h, followed by DAPI staining for 10 min. All incubations were done at room temperature. Finally, coverslips were mounted on slides with mounting media (Abcam, ab128982), imaged with a spinning disk confocal microscope (Quorum Technologies) using the same acquisition setting across all imaging sessions, and analyzed with Volocity software (v6.5.1). Protein levels are measured using fluorescence intensity and expressed relative to each other (e.g. ratios in Fig 6C) or relative to fluorescence intensity for YFP in the same cells. Each condition was tested in a minimum of 3 animals.
+
+
+Behavioral testing
+
Behavioral tests were performed on adult mice (male and female, 8-12 weeks). Mice were acclimated to the testing environment for at least 1 h the day prior to start of experiments. Behavioral testing (von Frey test and Hargreaves test) were then performed for 2-3 consecutive days for baseline and for another 3 days after CFA injection. Behavioral tests were performed at the same time in the morning, at room temperature (21°C) following a one hour acclimation period. Animals were randomly assigned to experimental groups and the experimenter was blind to the drug condition.
+
+CFA injection
+
CFA (Sigma, F5881) was thoroughly dissolved in saline (1:1) by vortexing the mixture. The resulting CFA solution (20 µl) was injected subdermally into the left hind paw under light isoflurane anaesthesia. The injection was performed shortly after the last baseline test, on Day 0.
+
+
+PF-71 administration
+
Injectable PF-71 solution was prepared by first dissolving PF-71 in DMSO to make a 5% stock solution; dissolution was achieved by heating to 37°C and vortexing. On the day of injection, stock solution was dissolved in sunflower oil (5% v/v) by sonicating for 5 min. Freshly prepared final PF-71 solution was injected intraperitoneally (1 g/kg body weight). Behavioral testing was performed 2 hours after injection of PF-71 or vehicle.
+
+
+Von Frey testing
+
Mechanical hyperalgesia was assessed with von Frey filaments (North Coast) using the SUDO method (93) . The average of 3 trials in each animal was used for analysis.
+
+
+Hargreaves testing
+
Thermal hyperalgesia was assessed with the Hargreaves apparatus (Ugo Basile, Italy). Radiant heat was applied to the plantar surface of the left hind paw. Interval between stimulus onset and paw withdrawal was defined as paw withdrawal latency (PWL). A 20 s cut-off prevented damage to the skin if the animal failed to withdraw. The average of 3 trials in each animal was used for analysis.
+
+
+
+Statistical analysis
+
Results are expressed as mean ± SEM when data are normally distributed or otherwise as median and quartiles. Normality was tested using the Kolmogorov-Smirnov test. Analysis was performed with GraphPad Prism (v9) and SigmaPlot (v11). Normally distributed data were compared using t-tests or two-way ANOVAs followed by a Student Newman-Keuls post hoc test. Non-normally distributed data were compared using Mann-Whitney and Wilcoxon signed rank tests. Fisher exact and McNemar test were used for categorical data. Exact significance values and test results are reported throughout figure legends.
+
+
+Computer model
+
Two separate, single compartmental models were built for DIV 0 and DIV4-7. They have the same, seven conductances: (𝑔̄Nav1.3, 𝑔Nav1.7, 𝑔Nav1.8, 𝑔Kdr, 𝑔M, 𝑔AHP and 𝑔Leak. Channel equations and their gating parameters are provided in Supplementary Table 4. Conductance densities at baseline (Supplementary Table 5) were tuned to qualitatively reproduce the changes in NaV channel expressions at DIV 0 and 4-7 indicated by the experiments; changes to other channels were minimized between the two models. The effects of ICA, PF-71 and PF-24 were simulated by adjusting (𝑔̄Nav1.3, 𝑔̄Nav1.7 and 𝑔̄Nav1.8=, respectively, as reported in the figures. Channel noise was added as Ornstein-Uhlenbeck process with 𝜇56782= 0 µA/cm>, 𝜎56782= 0.05 µA/cm>, and 𝜏 = 5 ms. All computer code is available at ModelDB (http://modeldb.yale.edu/267560; password: excitability).
+
CONTRIBUTIONS: YX, SR, SAP designed the research; YX, JY collected data, YX, JY, SR, SAP analyzed data; YX, SR, SAP wrote the manuscript.
+
ACKNOWLEDGMENTS: This work was supported by a Restracomp fellowship to JY and by a Canadian Institutes of Health Research Foundation Grant (FDN167276) to SAP. We thank Rohini Kuner for providing Nav1.8-Cre mice, Jason Jeong and Russell Smith for expert technical assistance with animal care and cell cultures, and Yongqian Wang for advice on qPCR data acquisition.
+
+
+
+
+
+REFERENCES
+CohenSP, VaseL, & HootenWM (2021) Chronic pain: an update on burden, best practices, and new advances. Lancet397():2082–2097.
+FinnerupNB, et al. (2015) Pharmacotherapy for neuropathic pain in adults: a systematic review and meta-analysis. Lancet Neurol14():162–173.
+RosenbergerDC, BlechschmidtV, TimmermanH, WolffA, & TreedeRD (2020) Challenges of neuropathic pain: focus on diabetic neuropathy. J Neural Transm127():589–624.
+HayM, ThomasDW, CraigheadJL, EconomidesC, & RosenthalJ (2014) Clinical development success rates for investigational drugs. Nat Biotechnol32():40–51.
+WoolfCJ (2010) Overcoming obstacles to developing new analgesics. Nat Med16():1241–1247.
+MogilJS (2009) Animal models of pain: progress and challenges. Nat Rev Neurosci10():283–294.
+TanejaA, Di IorioVL, DanhofM, & Della PasquaO (2012) Translation of drug effects from experimental models of neuropathic pain and analgesia to humans. Drug Discov Today17():837–849.
+MaoJ (2012) Current challenges in translational pain research. Trends Pharmacol Sci33():568–573.
+EdelmanGM & GallyJA (2001) Degeneracy and complexity in biological systems. Proc Natl Acad Sci U S A98():13763–13768.
+RattéS & PrescottSA (2016) Afferent hyperexcitability in neuropathic pain and the inconvenient truth about its degeneracy. Curr Opin Neurobiol36:31–37.
+GoldMS & GebhartGF (2010) Nociceptor sensitization in pain pathogenesis. Nat Med16():1248–1257.
+HaroutounianS, et al. (2014) Primary afferent input critical for maintaining spontaneous pain in peripheral neuropathy. Pain155():1272–1279.
+YatzivSL & DevorM (2019) Suppression of neuropathic pain by selective silencing of dorsal root ganglion ectopia using nonblocking concentrations of lidocaine. Pain160():2105–2114.
+AllesSRA & SmithPA (2021) Peripheral voltage-gated cation channels in neuropathic pain and their potential as therapeutic targets. Front Pain Res2:750583.
+BeanBP (2007) The action potential in mammalian central neurons. Nat Rev Neurosci. 8():451–465.
+WaxmanSG & ZamponiGW (2014) Regulating excitability of peripheral afferents: emerging ion channel targets. Nat Neurosci17():153–163.
+CoxJJ, et al. (2006) An SCN9A channelopathy causes congenital inability to experience pain. Nature444():894–898.
+FertlemanCR, et al. (2006) SCN9A mutations in paroxysmal extreme pain disorder: allelic variants underlie distinct channel defects and phenotypes. Neuron52():767–774.
+YangY, et al. (2004) Mutations in SCN9A, encoding a sodium channel alpha subunit, in patients with primary erythermalgia. J Med Genet41():171–174.
+Dib-HajjSD, YangY, BlackJA, & WaxmanSG (2013) The NaV1.7 sodium channel: from molecule to man. Nat Rev Neurosci14():49–62.
+NassarMA, et al. (2004) Nociceptor-specific gene deletion reveals a major role for NaV1.7 (PN1) in acute and inflammatory pain. Proc Natl Acad Sci U S A101():12706–12711.
+NassarMA, LevatoA, StirlingLC, & WoodJN (2005) Neuropathic pain develops normally in mice lacking both NaV1.7 and NaV1.8. Mol Pain1():24.
+MinettMS, et al. (2014) Pain without nociceptors? NaV1.7-independent pain mechanisms. Cell Rep6():301–312.
+GrubinskaB, et al. (2019) Rat NaV1.7 loss-of-function genetic model: Deficient nociceptive and neuropathic pain behavior with retained olfactory function and intra-epidermal nerve fibers. Mol Pain15:1744806919881846.
+MinettMS, et al. (2012) Distinct Nav1.7-dependent pain sensations require different sets of sensory and sympathetic neurons. Nat Commun3:791.
+ShieldsSD, et al. (2018) Insensitivity to pain upon adult-onset deletion of NaV1.7 or its blockade with selective inhibitors. J Neurosci38():10180–10201.
+MacDonaldDI, et al. (2021) A central mechanism of analgesia in mice and humans lacking the sodium channel NaV1.7. Neuron109():1497–1512.e1496.
+MinettMS, et al. (2015) Endogenous opioids contribute to insensitivity to pain in humans and mice lacking sodium channel NaV1.7. Nat commun6:8967.
+DehenH, WillerJC, PrierS, BoureauF, & CambierJ (1978) Congenital insensitivity to pain and the “morphine-like” analgesic system. Pain5():351–358.
+EmeryEC, LuizAP, & WoodJN (2016) NaV1.7 and other voltage-gated sodium channels as drug targets for pain relief. Expert Opin Ther Targets20():975–983.
+VetterI, et al. (2017) NaV1.7 as a pain target - From gene to pharmacology. Pharmacol Ther172:73–100.
+YangY, MisMA, EstacionM, Dib-HajjSD, & WaxmanSG (2018) NaV1.7 as a pharmacogenomic target for pain: Moving toward precision medicine. Trends Pharmacol Sci39():258–275.
+KushnarevM, PirvulescuIP, CandidoKD, & KnezevicNN (2020) Neuropathic pain: preclinical and early clinical progress with voltage-gated sodium channel blockers. Expert Opin Investig Drugs29():259–271.
+AlsaloumM, HigerdGP, EffraimPR, & WaxmanSG (2020) Status of peripheral sodium channel blockers for non-addictive pain treatment. Nat Rev Neurol16():689–705.
+EaglesDA, ChowCY, & KingGF (2022) Fifteen years of NaV 1.7 channels as an analgesic target: Why has excellent in vitro pharmacology not translated into in vivo analgesic efficacy?Br J Pharmacol179():3592–3611.
+KitanoY & ShinozukaT (2022) Inhibitio of NaV1.7: the possibility of ideal analgesics. RSC Med Chem13():895–920.
+MulcahyJV, et al. (2019) Challenges and Opportunities for Therapeutics Targeting the Voltage-Gated Sodium Channel Isoform NaV1.7. J Med Chem62():8695–8710.
+BankarG, et al. (2018) Selective NaV1.7 antagonists with long residence time show improved efficacy against inflammatory and neuropathic pain. Cell Rep24():3133–3145.
+KingwellK (2019) NaV1.7 withholds its pain potential. Nat Rev Drug Discov18:321–323.
+AlexandrouAJ, et al. (2016) Subtype-selective small molecule inhibitors reveal a fundamental role for NaV1.7 in nociceptor electrogenesis, axonal conduction and presynaptic release. PLoS One11():e0152405.
+McDermottLA, et al. (2019) Defining the functional role of NaV1.7 in human nociception. Neuron101():905–919.e908.
+RothenbergME, et al. (2019) Safety, tolerability, and pharmacokinetics of GDC-0276, a novel NaV1.7 Inhibitor, in a first-in-human, single- and multiple-dose study in healthy volunteers. Clin Drug Investig39():873–887.
+FlakeNM, LancasterE, WeinreichD, & GoldMS (2004) Absence of an association between axotomy-induced changes in sodium currents and excitability in DRG neurons from the adult rat. Pain109():471–480.
+ZhangJM, DonnellyDF, SongXJ, & LaMotteRH (1997) Axotomy increases the excitability of dorsal root ganglion cells with unmyelinated axons. J Neurophysiol. 78():2790–2794.
+AmirR, MichaelisM, & DevorM (1999) Membrane potential oscillations in dorsal root ganglion neurons: role in normal electrogenesis and neuropathic pain. J Neurosci. 19():8589–8596.
+CaffreyJM, EngDL, BlackJA, WaxmanSG, & KocsisJD (1992) Three types of sodium channels in adult rat dorsal root ganglion neurons. Brain Res592():283–297.
+RenganathanM, CumminsTR, & WaxmanSG (2001) Contribution of NaV1.8 sodium channels to action potential electrogenesis in DRG neurons. J Neurophysiol86():629–640.
+RushAM, CumminsTR, & WaxmanSG (2007) Multiple sodium channels and their roles in electrogenesis within dorsal root ganglion neurons. J Physiol579():1–14.
+ZhangJM, SongXJ, & LaMotteRH (1999) Enhanced excitability of sensory neurons in rats with cutaneous hyperalgesia produced by chronic compression of the dorsal root ganglion. J Neurophysiol82():3359–3366.
+PayneCE, et al. (2015) A novel selective and orally bioavailable Nav 1.8 channel blocker, PF-01247324, attenuates nociception and sensory neuron excitability. Br J Pharmacol172():2654–2670.
+VijayaragavanK, O’LearyME, & ChahineM (2001) Gating properties of NaV1.7 and NaV1.8 peripheral nerve sodium channels. J Neurosci21():7909–7918.
+TheileJW, FullerMD, & ChapmanML (2016) The selective NaV1.7 inhibitor, PF-05089771, interacts equivalently with fast and slow inactivated NaV1.7 channels. Mol Pharmacol90():540–548.
+McCormackK, et al. (2013) Voltage sensor interaction site for selective small molecule inhibitors of voltage-gated sodium channels. Proc Natl Acad Sci U S A110():E2724–2732.
+StregePR, et al. (2017) Sodium channel NaV1.3 is important for enterochromaffin cell excitability and serotonin release. Sci Rep7():15650.
+BennettDL, ClarkAJ, HuangJ, WaxmanSG, & Dib-HajjSD (2019) The role of voltage-gated sodium channels in pain signaling. Physiol Rev99():1079–1151.
+PrescottSA, De KoninckY, & SejnowskiTJ (2008) Biophysical basis for three distinct dynamical mechanisms of action potential initiation. PLoS Comput.Biol. 4():e1000198.
+BertaT, et al. (2008) Transcriptional and functional profiles of voltage-gated Na+ channels in injured and non-injured DRG neurons in the SNI model of neuropathic pain. Mol Cell Neurosci37():196–208.
+DustrudeET, WilsonSM, JuW, XiaoY, & KhannaR (2013) CRMP2 protein SUMOylation modulates NaV1.7 channel trafficking. J Biol Chem288():24316–24331.
+YamaneM, et al. (2017) A functional coupling between CRMP1 and NaV1.7 for retrograde propagation of Semaphorin3A signaling. J Cell Sci130():1393–1403.
+GouldHJ, 3rd, EnglandJD, LiuZP, & LevinsonSR (1998) Rapid sodium channel augmentation in response to inflammation induced by complete Freund’s adjuvant. Brain Res802():69–74.
+BlackJA, LiuS, TanakaM, CumminsTR, & WaxmanSG (2004) Changes in the expression of tetrodotoxin-sensitive sodium channels within dorsal root ganglia neurons in inflammatory pain. Pain108():237–247.
+LiangL, FanL, TaoB, YasterM, & TaoYX (2013) Protein kinase B/Akt is required for complete Freund’s adjuvant-induced upregulation of NaV1.7 and NaV1.8 in primary sensory neurons. J Pain14():638–647.
+AkinEJ, et al. (2019) Building sensory axons: Delivery and distribution of NaV1.7 channels and effects of inflammatory mediators. Sci Adv5():eaax4755.
+MorenoAM, et al. (2021) Long-lasting analgesia via targeted in situ repression of NaV1.7 in mice. Sci Transl Med13(584).
+BaronR & DickensonAH (2014) Neuropathic pain: precise sensory profiling improves treatment and calls for back-translation. Pain155():2215–2217.
+MarderE & GoaillardJM (2006) Variability, compensation and homeostasis in neuron and network function. Nat Rev Neurosci7():563–574.
+O’LearyT, WilliamsAH, FranciA, & MarderE (2014) Cell types, network homeostasis, and pathological compensation from a biologically plausible ion channel expression model. Neuron82():809–821.
+DrionG, O’LearyT, & MarderE (2015) Ion channel degeneracy enables robust and tunable neuronal firing rates. Proc Natl Acad Sci U S A112():E5361–5370.
+MishraP & NarayananR (2022) Conjunctive changes in multiple ion channels mediate activity-dependent intrinsic plasticity in hippocampal granule cells. iScience25():103922.
+YangJ, ShakilH, RattéS, & PrescottSA (2022) Minimal requirements for a neuron to coregulate many properties and the implications for ion channel correlations and robustness. Elife11:e72875.
+DrionG, MassotteL, SepulchreR, & SeutinV (2011) How modeling can reconcile apparently discrepant experimental results: the case of pacemaking in dopaminergic neurons. PLoS Comput Biol7():e1002050.
+PuopoloM, RaviolaE, & BeanBP (2007) Roles of subthreshold calcium current and sodium current in spontaneous firing of mouse midbrain dopamine neurons. J Neurosci27():645–656.
+SwensenAM & BeanBP (2005) Robustness of burst firing in dissociated purkinje neurons with acute or long-term reductions in sodium conductance. J Neurosci25():3509–3520.
+KnoxAT, GlauserT, TenneyJ, LyttonWW, & HollandK (2018) Modeling pathogenesis and treatment response in childhood absence epilepsy. Epilepsia59():135–145.
+PrinzAA, BucherD, & MarderE (2004) Similar network activity from disparate circuit parameters. Nat Neurosci7():1345–1352.
+GrashowR, BrookingsT, & MarderE (2010) Compensation for variable intrinsic neuronal excitability by circuit-synaptic interactions. J Neurosci30():9145–9156.
+MedlockL, et al. (2022) Multiscale computer model of the spinal dorsal horn reveals changes in network processing associated with chronic pain. J Neurosci42():3133–3149.
+PrescottSA, SejnowskiTJ, & De KoninckY (2006) Reduction of anion reversal potential subverts the inhibitory control of firing rate in spinal lamina I neurons: towards a biophysical basis for neuropathic pain. Mol.Pain2.
+LeeKY & PrescottSA (2015) Chloride dysregulation and inhibitory receptor blockade yield equivalent disinhibition of spinal neurons yet are differentially reversed by carbonic anhydrase blockade. Pain156():2431–2437.
+DjouhriL, et al. (2003) Sensory and electrophysiological properties of guinea-pig sensory neurones expressing NaV1.7 (PN1) Na+ channel alpha subunit protein. J Physiol546():565–576.
+CampbellJN, RajaSN, MeyerRA, & MackinnonSE (1988) Myelinated afferents signal the hyperalgesia associated with nerve injury. Pain32():89–94.
+KoltzenburgM, LundbergLE, & TorebjorkHE (1992) Dynamic and static components of mechanical hyperalgesia in human hairy skin. Pain51():207–219.
+LiuCN, et al. (2000) Tactile allodynia in the absence of C-fiber activation: altered firing properties of DRG neurons following spinal nerve injury. Pain85():503–521.
+LiuX, EschenfelderS, BlenkKH, JänigW, & HäblerH (2000) Spontaneous activity of axotomized afferent neurons after L5 spinal nerve injury in rats. Pain84():309–318.
+AbrahamsenB, et al. (2008) The cell and molecular basis of mechanical, cold, and inflammatory pain. Science321():702–705.
+GoaillardJM & MarderE (2021) Ion channel degeneracy, variability, and covariation in neuron and circuit resilience. Annu Rev Neurosci44:335–357.
+YangJ & PrescottSA (2023) Homeostatic regulation of neuronal function: importance of degeneracy and pleiotropy. Front Cell Neurosci17:1184563.
+SchildJH & KunzeDL (1997) Experimental and modeling study of Na+ current heterogeneity in rat nodose neurons and its impact on neuronal discharge. J Neurophysiol78():3198–3209.
+BlairNT & BeanBP (2002) Roles of tetrodotoxin (TTX)-sensitive Na+ current, TTX-resistant Na+ current, and Ca2+ current in the action potentials of nociceptive sensory neurons. J Neurosci22():10277–10290.
+AgarwalN, OffermannsS, & KunerR (2004) Conditional gene deletion in primary nociceptive neurons of trigeminal ganglia and dorsal root ganglia. Genesis38():122–129.
+MalinSA, DavisBM, & MolliverDC (2007) Production of dissociated sensory neuron cultures and considerations for their use in studying neuronal function and plasticity. Nat Protoc2():152–160.
+DavidsonS, et al. (2014) Human sensory neurons: Membrane properties and sensitization by inflammatory mediators. Pain155():1861–1870.
+BoninRP, BoriesC, & De KoninckY (2014) A simplified up-down method (SUDO) for measuring mechanical nociception in rodents using von Frey filaments. Mol Pain10:26.
+
+
+
+
+
Inhibiting Na<sub>V</sub>1.8 at DIV0 with A-803467 had the same effect as PF-24.
(A) Sample voltage clamp recording at DIV0 before and after A-803467 (1 µM). (B) Peak current was significantly reduced by A-803467 (F1,84=9.935, p=0.016, two-way RM ANOVA, n=8). *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc test in B.
+
+
+
+
+
Inhibiting Na<sub>V</sub>1.8 with PF-24 at DIV4-7 had negligible effects.
(A) Inhibiting Nav1.8 with PF-24 (1 µM) did not affect spiking pattern (χ2 =0.00, p=1.00, McNemar test) and modestly reduced firing rate (F1,54=9.745, p= 0.012, two-way RM ANOVA, n=10) in DIV4-7 neurons. (B) PF-24 did not affect rheobase (Z12=0.420, p=0.685, Wilcoxon Rank test) but did reduce spike height (T12=2.939, p=0.012, paired-t-test). *, p<0.05; **, p<0.01; Student-Newman-Keuls post-hoc tests in A.
+
+
+
+
+
Inhibiting Na<sub>V</sub>1.7 at DIV0 had negligible effects.
(A) Inhibiting NaV1.7 with PF-71 (30 nM) did not alter spiking pattern (χ2=0.00, p=1.00, McNemar test) or reduce firing rate (F1,30=5.805, p=0.061, two-way RM ANOVA, n=6) in DIV0 neurons; in fact, firing rate was slightly increased. (D-E) PF-71 did not affect rheobase (Z9=0.677, p=0.578, Wilcoxon rank test) but did reduce spike height (T9=3.759, p=0.004, paired-t-test).
+
+
+
+
+
Na<sub>V</sub>1.7 and Na<sub>V</sub>1.3 currents can compensate for each other.
(A) In the DIV4-7 model, reducing 𝑔̄Nav1.7 to 30% of its normal value (35®10 mS/cm2) can be compensated for by increasing 𝑔̄Nav1.3 to 229% of its normal value (0.35®0.8 mS/cm2) to maintain repetitive spiking. (B) Conversely, reducing 𝑔̄Nav1.3 to 10% of its normal value (0.35®0.035 mS/cm2) can be compensated for by increasing 𝑔̅Nav1.7 to 171% of its normal value (35®60 mS/cm2) and maintain repetitive spiking.
+
+
+
+
+
Virtual conductances restored repetitive spiking after pharmacological inhibition of the corresponding native conductance had converted the neuron to transient spiking.
(A) Sample response at DIV0 showing that a virtual NaV1.8 conductance applied with dynamic clamp restored repetitive spiking after inhibiting native NaV1.8 channels with PF-24. This restoration was repeated in 3 of 3 neurons tested. (B) Sample recording at DIV4-7 showing that a virtual NaV1.7 conductance restored repetitive spiking after inhibiting native NaV1.7 channels with PF-71. This restoration was repeated in 4 of 4 neurons tested.
+
+
+
+
+
Channel inactivation affects Na<sub>V</sub> subtype contribution on short timescale.
(A) In the DIV0 model, NaV1.7 contributed to the first spike but its inactivation meant that all subsequent spikes relied exclusively on NaV1.8. (B) In the DIV4-7 model, despite some inactivation of NaV1.3 (maroon) and NaV1.7 (orange), the remaining current was still large enough (because of the higher gmax of those two subtypes) to produce inward current sufficient to support repetitive spiking despite the low gmax of NaV1.8 in the DIV4-7 model.
+
+
+
+
+
Effect of TTX on voltage threshold in DIV0 neurons.
Despite TTX having negligible effects in DIV0 neurons according to our initial analysis (see Fig. 1), simulation results in Fig. 5A,B predicted that the first spike was nonetheless initiated by NaV1.7. By extension, this predicted that TTX should cause a depolarizing shift in voltage threshold for the first spike. Analysis of the experimental data confirmed this to be true, with threshold (mean±SEM) increasing from -33.7±1.4 mV at baseline to -28.3±1.4 mV after TTX (T24=-3.19, p=0.004, paired t-test). Confirmation of this unexpected prediction helps further validate our model neuron.
+
+
+
+
Model data before and after channel “inhibition”
+
+
+
+
Reagents
+
+
+
+
Primers
+
+
+
+
Model equations
+
+
+
+
+
Conductance densities at baseline for DIV 0 and 4-7 models
+
+
+
+
+
diff --git a/test/fixtures/2023.06.26.546606/2023.06.26.546606.xml b/test/fixtures/2023.06.26.546606/2023.06.26.546606.xml
new file mode 100644
index 00000000..f4c68da4
--- /dev/null
+++ b/test/fixtures/2023.06.26.546606/2023.06.26.546606.xml
@@ -0,0 +1,549 @@
+
+
+
+
+
+BIORXIV
+
+bioRxiv
+bioRxiv
+
+
+Cold Spring Harbor Laboratory
+
+
+
+10.1101/2023.06.26.546606
+1.1
+
+
+Regular Article
+
+
+New Results
+
+
+Cancer Biology
+
+
+
+Metabolic reprogramming of cancer cells by JMJD6-mediated pre-mRNA splicing is associated with therapeutic response to splicing inhibitor
+
+
+
+JablonowskiCarolyn
+1
+#
+
+
+QuarniWaise
+1
+#
+
+
+SinghShivendra
+1
+
+
+TanHaiyan
+2
+
+
+BostanthirigeDhanushka Hewa
+1
+
+
+JinHongjian
+3
+
+
+FangJie
+1
+
+
+ChangTi-Cheng
+2
+
+
+FinkelsteinDavid
+3
+
+
+ChoJi-Hoon
+2
+
+
+HuDongli
+1
+
+
+PagalaVishwajeeth
+2
+
+
+SakuradaSadie Miki
+4
+
+
+Pruett-MillerShondra M.
+4
+
+
+WangRuoning
+5
+
+
+MurphyAndrew
+1
+
+
+FreemanKevin
+6
+
+
+PengJunmin
+2
+9
+
+
+DavidoffAndrew M
+1
+7
+8
+
+
+http://orcid.org/0000-0002-1678-5864
+WuGang
+3
+
+
+YangJun
+1
+7
+8
+9
+
+Department of Surgery, St Jude Children’s Research Hospital, Memphis, TN38105, USA
+Center for Proteomics and Metabolomics, Department of Structural Biology, Department of Developmental Neurobiology, St Jude Children’s Research Hospital, Memphis, TN38105, USA
+Center for Applied Bioinformatics, St Jude Children’s Research Hospital, Memphis, TN38105, USA
+Department of Cell and Molecular Biology, St. Jude Children’s Research Hospital, Memphis, TN 38105, USA
+Center for Childhood Cancer and Blood Disease, Abigail Wexner Research Institute, Nationwide Children’s Hospital, 700 Children’s Drive, Columbus, OH 43205, USA
+Genetics, Genomics & Informatics, The University of Tennessee Health Science Center (UTHSC), Memphis, TN 38103, USA
+St Jude Graduate School of Biomedical Sciences, St Jude Children’s Research Hospital, TN 38105
+Department of Pathology, College of Medicine, The University of Tennessee Health Science Center, 930 Madison Ave, Suite 500, Memphis, TN 38163
+
+
+
equal contribution
+Correspondence: Jun.Yang2@stjude.org
+
The authors declare no potential conflicts of interest
Dysregulated pre-mRNA splicing and metabolism are two hallmarks of MYC-driven cancers. Pharmacological inhibition of both processes has been extensively investigated as potential therapeutic avenues in preclinical and clinical studies. However, how pre-mRNA splicing and metabolism are orchestrated in response to oncogenic stress and therapies is poorly understood. Here, we demonstrate that JMJD6 acts as a hub connecting splicing and metabolism in MYC-driven neuroblastoma. JMJD6 cooperates with MYC in cellular transformation by physically interacting with RNA binding proteins involved in pre-mRNA splicing and protein homeostasis. Notably, JMJD6 controls the alternative splicing of two isoforms of glutaminase (GLS), namely kidney-type glutaminase (KGA) and glutaminase C (GAC), which are rate-limiting enzymes of glutaminolysis in the central carbon metabolism in neuroblastoma. Further, we show that JMJD6 is correlated with the anti-cancer activity of indisulam, a “molecular glue” that degrades splicing factor RBM39, which complexes with JMJD6. The indisulam-mediated cancer cell killing is at least partly dependent on the glutamine-related metabolic pathway mediated by JMJD6. Our findings reveal a cancer-promoting metabolic program is coupled with alternative pre-mRNA splicing through JMJD6, providing a rationale to target JMJD6 as a therapeutic avenue for treating MYC-driven cancers.
+
+
+
+
+
+
+
+Competing Interest Statement
The authors have declared no competing interest.
+
+
+
+
+Introduction
+
Metabolic reprogramming is a hallmark of cancer1-3 which allows rapidly proliferating tumor cells to acquire nutrients to meet their bioenergetic, biosynthetic, and redox demands4. One of the primary driving forces in reprograming cancer cell metabolism is the deregulated MYC family proto-oncogenes (C-MYC, MYCN, and MYCL)5, which are known to encode master transcriptional factors that regulate metabolic gene expression. MYC coordinates nutrient acquisition to produce ATP and key cellular building blocks that increase cell mass and promote DNA replication and cell division6. The increase in total RNA and protein synthesis by overactive MYC signaling leads to dysregulation of macromolecular processing machineries including the spliceosome7, and consequently pre-mRNA splicing8-10, another hallmark of MYC-driven cancers7,9-12, for the purpose of cellular stress adaptation. MYCN amplification is one of the most important biological features of high-risk neuroblastoma13. Transgenic mouse and zebrafish models have demonstrated that MYCN is a neuroblastoma driver14,15. In tumors without MYCN amplification, C-MYC is overexpressed, further indicating that neuroblastoma is a MYC-driven cancer. The metabolic dependency of neuroblastoma has been widely studied by us and others16-23. A larger number of splicing changes have also been noticed in high-stage neuroblastomas24-26. Splicing alterations lead to a spliceosomal vulnerability that provides a new opportunity to develop transformative therapies by disrupting aberrant pre-mRNA splicing7,9-12. We and others have shown that targeting the splicing factor RBM39 by indisulam, a “molecular glue” that bridges RBM39 to E3 ubiquitin ligase DCAF15 for proteasomal degradation, achieved an exceptional anti-tumor activity in neuroblastoma models27,28. Disruption of spliceosome by Pladienolide B also resulted in significant anti-tumor effect in neuroblastoma models26 However, how the dysregulated pre-mRNA splicing machinery and metabolism are orchestrated in MYC-driven neuroblastoma has not been well elucidated. Whether metabolism modulates the anti-cancer effect of splicing inhibition remains to be answered.
+
Next-generation sequencing studies have revealed only a few recurrent somatic mutations in neuroblastoma at the time of diagnosis29,30. However, copy number alterations of chromosomal segments such as 17q gain, 1p36 or 11q23 loss frequently occur in high-risk neuroblastoma. While attempts to understand the functions of individual genes in these chromosomal segments have been reported (i.e., BIRC531, PHB32, PPM1D33, TRIM3734 in 17q; ARID1A35, CAMTA136, CASZ137, CHD538-40, KIF1Bβ41-43, miR-34a44, RUNX345 in 1p36), the biological consequences of these genetic events in MYC-driven tumors still remain largely unknown. Gain of 17q is the most frequent genetic event in high-risk neuroblastoma and is associated with MYCN amplification46. In addition, in the transgenic MYCN mouse model of neuroblastoma, the chromosomal locus syntenic to human 17q is partially amplified47, indicating that chromosome 17q is needed for MYC-mediated tumorigenesis.
+
JMJD6 is a JmjC domain–containing nuclear protein with iron- and 2-oxoglutarate–dependent dioxygenase activity48, whose coding gene is located on chromosome 17q25. While the histone arginine demethylase activity of JMJD6 that catalyzes demethylation of H4R3me1/me2 is controversial49, JMJD6 is a lysyl-5-hydroxylase that catalyzes 5-hydroxylation on specific lysine residues of target proteins50. JMJD6 has pleiotropic functions in normal physiology and in cancer51-54. We previously found that JMJD6 is essential for the survival of neuroblastoma cells (including MYCN-amplified and C-MYC– overexpressed cells)55, which was further validated by an independent study56, indicating that neuroblastoma has JMJD6 dependency. However, the exact mechanism of JMJD6 in MYC-driven cancers remains elusive. One study has shown that JMJD6 and BRD4 co-bind at antipause enhancers, regulating promoter-proximal pause release of a large subset of transcription units57. By harnessing a similar mechanism, JMJD6 promotes cell survival of glioblastoma in vivo58. These findings are particularly interesting because BRD4 occupies exceptionally large super-enhancers associated with genes, including C-MYC and MYCN59-61, and the expression of those enhancers can be disrupted by BRD4 inhibitors, which have a potent antitumor effect59-62. Here we show a new mechanism by which JMJD6 promotes tumorigenesis mediated by the MYC oncogene in that JMJD6 interacts with a subset of RNA binding proteins including RBM39 in neuroblastoma cells and regulates the alternative splicing of metabolic genes that are involved in mitochondrial metabolism. “Glutamine addiction” is one key feature of MYC-driven tumors. Glutaminase (GLS) is the enzyme responsible for conversion of glutamine to glutamate in the process of glutaminolysis to feed the tricarboxylic acid (TCA) cycle and has two splice isoforms, GAC (glutaminase C) and KGA (kidney-type glutaminase). We show that JMJD6 controls the alternative splicing of KGA and GAC, and, consequently, impacts the central carbon metabolism in neuroblastoma. Further we show that JMJD6 is correlated with the anti-cancer activity of indisulam, a “molecular glue” that degrades the splicing factor RBM39. The indisulam-mediated cancer cell killing is at least partly dependent on the glutamine-related metabolic pathway mediated by JMJD6. Our findings demonstrate a new mechanism by which JMJD6 coordinates metabolic programs and alternative pre-mRNA splicing, providing a rationale to target JMJD6 as a therapeutic target for MYC-driven cancers.
+
+
+Results
+
+The essential genes for neuroblastoma cell survival on chromosome 17q target pre-mRNA splicing and metabolism
+
An incomplete understanding of the biological consequences of chromosome 17q gain remains a barrier to the understanding of high-risk neuroblastoma. 1132 genes are located on 17q (Supplementary table 1). We surmised that some of the 17q genes are particularly important for neuroblastoma cell survival. Analysis of the cancer dependency genes in neuroblastoma cell lines screened with the Avana sgRNA library63 revealed that 114 were essential to neuroblastoma (mean score <-0.4) (Figure 1a, Supplementary table 2). Protein interaction network analysis followed by functional annotation revealed that proteins encoded by these 114 essential genes formed distinct but interconnected modules including RNA splicing (i.e., SRSF2, DDX5, DDX42, DHX8), mitochondrial metabolism (i.e., NDUFA8, COX11, SLC25A10, SLC35B1), protein homeostasis (i.e., UBE2O, PSMB3, PSMC5), DNA repair (i.e., BRIP1, BRCA1, RAD51C, RAD51D) and transcriptional regulation (i.e., PHF12, CBX1, SMARCE1, MED1), as well as endocytosis (i.e., CHMP6, CTLC, EPN3, HGS, SNF8, VPS25) (Figure 1b). Using these 114 genes as a signature, we found that 81 of them were highly expressed in high-risk neuroblastomas, which were enriched with MYCN amplification (Figure 1c). Correspondingly, neuroblastomas with high expression levels of this gene signature were associated with a poorer event-free and overall survival of patients (Figure 1d, e). These data demonstrate that 17q genes are involved in essential biological processes and highly expressed in high-risk neuroblastomas.
+
+
+
17q contains neuroblastoma dependency genes
+
a. CRISPR score for 17q genes in 10 neuroblastoma cell lines. Score <-0.4 is defined as neuroblastoma dependency genes. Data are derived from Avana sgRNA library screening63.
+
b. STRING protein interaction network showing 17q essential genes with various biological functions.
+
c. Heatmap by K-means clustering analysis showing 17q essential genes are highly expressed in high-risk neuroblastomas based on RNA-seq data (SEQC dataset).
+
d. Kaplan-Meier survival curve showing 17q essential gene signature is correlated with worse event-free survival (SEQC dataset).
+
e. Kaplan-Meier survival curve showing 17q essential gene signature is correlated with worse overall survival (SEQC dataset).
+
+
+
+
+
+JMJD6 is required for neuroblastoma growth
+
JMJD6 was among these 114 essential genes. To understand the role of JMJD6, we examined the genetic features of JMJD6 in neuroblastoma and other types of cancers. Among the genes encoding JmjC-domain containing proteins, JMJD6 was the only one that was frequently amplified in neuroblastoma (Figure 2a). High JMJD6 expression was associated with poor event-free outcome, as shown by Kaplan-Meier analysis (Figure 2b). To examine whether JMJD6 amplification is limited to specific tumor types, we explored genomic data from different cancers using the cBioportal program. JMJD6 was amplified across multiple types of adult cancers such as breast and liver cancer (Supplementary Figure 1a), and correlated with worse relapse-free survival (Supplementary Figure 1b). We further compared the RNA-seq expression of JMJD6 in 2337 samples across over 20 pediatric cancer subtypes and found that JMJD6 showed the highest expression levels in neuroblastoma (Supplementary Figure 1c), suggesting that JMJD6 might be particularly important in neuroblastoma. We validated this hypothesis using shRNA knockdown of JMJD6 in MYCN amplified cells (BE2C, SIMA, KELLY, IMR32) and non-MYCN amplified cells (SK-N-AS and CHLA20). The results showed that loss of JMJD6 greatly reduced the colony numbers in all tested cell lines (Figure 2c), demonstrating that JMJD6 is essential to neuroblastoma cells regardless of MYCN amplification. Neuroblastic tumors comprise a histologic spectrum that ranges from less-differentiated neuroblastoma to well-differentiated ganglioneuroma. The extent of differentiation in the tumor cells is correlated with prognostic significance64. We noticed that the loss of JMJD6 led to neurite outgrowth (Supplementary Figure 1d), a unique structure of neuroblastoma cells differentiating in vitro. This morphologic change suggests that JMJD6 is required to maintain cancer cell stemness. Lastly, we validated that JMJD6 is essential to neuroblastoma growth in MYCN amplified (BE2C) and C-MYC overexpressed (SK-N-AS) xenograft models (Figure 2d, e). Taken together, these data demonstrate that loss of JMJD6 function impedes neuroblastoma cell survival and tumor growth.
+
+
+
JMJD6 is required for neuroblastoma growth and facilitates MYC-mediated cellular transformation.
+
a. Copy number of genes encoding JmjC domain proteins in St Jude neuroblastoma cohort.
+
b. Kaplan-Meier survival curve showing high JMJD6 is correlated with worse event-free survival (SEQC RNA-seq dataset).
+
c. Crystal violet showing the colony staining after JMJD6 siRNA knockdown in neuroblastoma cell lines validated by western blot.
+
d. Xenograft tumor growth of BE2C (right) models with lentiviral JMJD6 shRNA knockdown. P-value calculated by multiple unpaired t-test across each row. n=5 per group.***p<0.001, **p<0.01.
+
e. Xenograft tumor growth of SK-N-AS models with lentiviral JMJD6 shRNA knockdown. P-value calculated by multiple unpaired t-test across each row. ***p<0.001, **p<0.01.
+
f. Western blot validating the expression of retroviral based MYCN and JMJD6 in JoMa1 cells.
+
g. Cell proliferation of JoMa1 cells transduced with indicated constructs, GFP, JMJD6, MYCN, JMJD6+MYCN.
+
h. Colony formation JoMa1 cells transduced with indicated constructs, GFP, JMJD6, MYCN, JMJD6+MYCN. Top panel showing photos taken under light microscope. Bottom panel showing cell colonies stained with crystal violet. P value calculated by multiple unpaired t-test across each row. ***p<0.001, **p<0.01.
+
i. Xenograft tumor growth of JoMa1 cells transduced with indicated constructs, GFP, JMJD6, MYCN, JMJD6+MYCN. n=5 per group. P value calculated by multiple unpaired t-test across each row. ***p<0.001, **p<0.01.
Next, we investigated whether gain of function of JMJD6 could facilitate MYC-mediated oncogenic transformation. To test this, we used a NIH3T3 transformation assay that provides a straightforward method to assess the transforming potential of an oncogene, which may lead to morphological transformation and loss of contact inhibition, a typical feature of cellular transformation. Like the GFP control (Supplementary Figure 2a), we found that NIH3T3 cells with overexpressed JMJD6 stopped proliferation after being confluent (Supplementary Figure 2b), indicating JMJD6 alone is unable to transform NIH3T3 cells. However, overexpression of MYCN induced foci formation with enhanced cell death (Supplementary Figure 2c). Importantly, NIH3T3 cells with overexpressed MYCN and JMJD6 lost contact inhibition, accompanied with morphological change, and formed larger foci (Supplementary Figure 2d), indicating that JMJD6 enhances MYCN activity to transform NIH3T3 cells. Interestingly, MYCN alone also reprogramed metabolism of NIH3T3 cells as shown by the color change of the media, indicative of acidic pH change, which was largely rescued by co-expression of JMJD6 (Supplementary Figure 2e), suggesting that cells with enhanced lactate production by MYCN were directed to oxidative phosphorylation by JMJD6. It is believed that the cell of origin of neuroblastoma is the progeny of neural crest cells. We therefore tested the role of JMJD6 in MYC-mediated transformation using JoMa1, a cell line derived from murine neural crest, by transducing GFP, JMJD6, MYCN and JMJD6/MYCN (Figure 2f). While JMJD6 showed no difference from GFP control in regulating cell proliferation, MYCN slightly enhanced cell proliferation (Figure 2g). However, the combination of JMJD6 and MYCN remarkably increased cell proliferation, mirrored by the colony formation assay which showed JMJD6/MYCN induced rapid growth of colonies with distinct transformation morphology (Figure 2g, h). Implantation of each group into immune-deficient mice led to tumor development of MYCN and JMJD6/MYCN groups (Figure 2i). However, JMJD6/MYCN group tumors appeared to grow faster than the MYCN alone tumors. Taken together, these data indicate that JMJD6 enhances the MYC-mediated transformation, demonstrating the oncogenic role of JMJD6.
+
+
+JMJD6 regulates pathways engaged in pre-mRNA splicing and mitochondrial biogenesis in neuroblastoma
+
We surmised that loss of function of genes in the same functional module/pathway may induce similar effects across different cell lineages, which in turn corroborates the hypothesis that JMJD6 is a player in that signaling pathway. To test this hypothesis, we analyzed the dependency correlation of JMJD6 knockout and other genes (defined as co-dependency genes if they are positively correlated), by using the DepMap data (https://depmap.org) that includes genome-wide knockout in 1027 cell lines of more than 20 cancer types (Supplementary Table 3), followed by pathway enrichment. The data showed that JMJD6 co-dependency genes were significantly and positively correlated with spliceosome/mRNA splicing (i.e., RBM39, SF3B1), ubiquitin-mediated proteolysis and endocytosis and a number of 17q25 genes, (Figure 3a, c), which mirrored the pathway network of 17q essential genes in neuroblastoma (Figure 1b). JMJD6 knockout was negatively correlated with the knockout of genes housed at chromosome 1p, which is frequently deleted in high-risk neuroblastoma, and oxidative phosphorylation as well as protein translation (Figure 3b, d).
+
+
+
JMJD6 regulates pre-mRNA splicing of genes involved in metabolism
+
a. Pathway enrichment for JMJD6 co-dependency genes whose knockout exhibits similar phenotype with JMJD6 knockout based on re-analysis of DepMAP data.
+
b. Pathway enrichment for genes whose knockout exhibits opposite phenotype with JMJD6 knockout based on re-analysis of DepMAP data.
+
c. Chromosomal location enrichment for JMJD6 co-dependency genes whose knockout exhibits similar phenotype with JMJD6 knockout based on re-analysis of DepMAP data.
+
d. Chromosomal location enrichment for genes whose knockout exhibits opposite phenotype with JMJD6 knockout based on re-analysis of DepMAP data.
+
e. Pathways analysis for genes downregulated and upregulated by JMJD6 knockdown commonly shared in SK-NAS and BE2C cells.
+
f. Alternative splicing events altered by JMJD6 knockdown in BE2C and SK-N-AS cells.
+
g. Pathway enrichment for each splicing event commonly shared in BE2C and SK-N-AS cells after JMJD6 knockdown.
+
h. Isoform identification based on splicing events in BE2C and SK-N-AS cells, followed by pathway enrichment for commonly shared alterations in both cell lines.
+
+
+
+
BRD4 is known to regulate MYC expression61,65. Previous studies have shown that JMJD6 and BRD4 interact to regulate gene transcription56,58,66, suggesting that JMJD6 might directly modulate MYC expression. To assess this possibility, we knocked down JMJD6 in neuroblastoma cell lines BE2(C) and SK-N-AS, which express MYCN and C-MYC, respectively, for RNA-seq analysis. The sequencing data were analyzed for differential gene expression (Supplementary Table 4). Interestingly, loss of JMJD6 showed minimal impact on expression of MYCN in BE2C cells or C-MYC in SK-N-AS cells (Supplementary Figure 3a), and western blot analysis did not show alteration of MYCN expression although C-MYC protein was slightly downregulated by loss of JMJD6 (Supplementary Figure 3b). However, BRD4 inhibitors drastically inhibited both MYCN and C-MYC expression in neuroblastoma cells59-61, suggesting that JMJD6 inhibition has a distinct effect from the BRD4 inhibition. Gene set enrichment analysis for pathway engagement for the genes commonly downregulated or upregulated in both cell lines revealed that loss of JMJD6 most significantly repressed the expression of genes involved in pre-mRNA splicing, histones, and cell cycle G1/S checkpoint (Figure 3e), and enhanced the pathways involved in mitochondrial functions and heat shock response (Figure 3e). Interestingly, the genes transcribed from the mitochondrial genome were elevated in both cell lines (Supplementary Figure 3c), suggesting that JMJD6 directly or indirectly regulates the transcription of mitochondrial genome. These data are consistent with the co-dependency pathways of JMJD6 (Figure 3a-d). Depletion of JMJD6 in both cell lines led to the downregulation of MYC signaling pathways (Supplementary Figure 4), although ranked behind the pathways of splicing and metabolism, suggesting that the MYC pathways are not primarily regulated by JMJD6. These data indicate that JMJD6 does not regulate the gene expression of the MYC family of transcription factors but might indirectly regulate the MYC pathway.
+
To verify if JMJD6 regulates pre-mRNA splicing, we analyzed the RNA-seq using two algorithms. The first one is event-based analysis to identify the altered exon splicing (Figure 3f). We found that knockdown of JMJD6 dominantly affects the exon skipping (SE) although other splicing events were also altered, albeit with a much smaller number (Figure 3f). Pathway analysis of common events in both BE2C and SK-N-AS cells demonstrated that genes involved in metabolism and splicing are the most significantly affected by loss of function of JMJD6 (Figure 3g). Using the second algorithm of RNA splicing analysis previously developed (Figure 3h) that allows discovery of new isoforms of genes generated through alternative splicing67, we identified 580 genes in BE2C cells and 1018 genes in SK-N-AS cells undergoing alternative splicing after JMJD6 knockdown, 133 of which were shared by both (Figure 3h, Supplementary Fig. 5, Supplementary Table 5). The alternatively spliced genes were involved in a variety of pathways (Supplementary Table 6). KEGG pathway enrichment analysis of the 133 commonly alternatively spliced genes showed that only metabolic pathway genes were significantly enriched (Figure 3f), most of which are involved in mitochondrial bioenergetics and folate metabolism (Figure 3h). Collectively, these data demonstrate that JMJD6 regulates RNA splicing of genes engaged in mitochondrial metabolism, being one of the key mediators of the 17q locus activity in neuroblastoma.
+
+
+JMJD6 regulates alternative splicing of glutaminolysis gene GLS
+
Overactive MYC signaling leads to altered macromolecular processing machineries in response to an increase in total RNA and protein synthesis7. MYC is also a master regulator of cancer metabolism involved in ribosomal and mitochondrial biogenesis, glucose and glutamine metabolism and lipid synthesis, leading to the acquisition of bioenergetic substrates enabling the cancer cell to grow and proliferate68-70. “Glutamine addiction” is one feature of MYC-driven cancer71. The pre-mRNA splicing altered by JMJD6 knockdown included GLS, the key enzyme of glutaminolysis, prompting us to investigate the GLS splicing mediated by JMJD6 knockdown. GLS is known to catalyze the conversion of glutamine to glutamate, and is alternatively spliced to form two isoforms, GAC and KGA72, with different cellular localization and catalytic capacities73. The GAC isoform is more frequently upregulated in cancer cells than KGA74, and has been shown to be regulated by MYC75-77, leading to a “glutamine addiction” phenotype in MYC-driven tumors71. We found that loss of JMJD6 led to a splicing switch from the GAC isoform (with exons 1-15) to the KGA isoform (with exons 1-14 and 16-19) (Figure 4a), which was further confirmed by isoform-specific RT PCR (Figure 4b). We then investigated the expression of GAC/KGA at the protein levels after JMJD6 knockdown. Western blot showed that the KGA isoform was increased after JMJD6 knockdown in all three tested cell lines, SK-N-AS, BE2C and SIMA (Figure 4c). Then we further validated the JMJD6 effect on GLS isoform expression by using a luciferase reporter that indicates the isoforms of GAC and KGA. Indeed, JMJD6 knockdown significantly increased the expression of the KGA reporter (Figure 4d). RNA-immunoprecipitation showed that JMJD6 bound to GLS RNA (Figure 4e), suggesting that JMJD6 may directly regulate the splicing of GLS. We reasoned that if the regulation of GLS splicing by JMJD6 was a bone fide mechanism, the expression levels of JMJD6 would correlate with the levels of GAC/KGA in tumors. Indeed, JMJD6 was positively correlated with GAC and negatively correlated with KGA in two independent neuroblastoma cohorts (Figure 4f), supporting the hypothesis that JMJD6 is required to maintain the high ratio of GAC/KGA in cancer cells by controlling their alternative splicing. Clinical relevance of GAC and KGA in neuroblastoma was evidenced by the findings that high GAC was associated with a worse event-free survival and high KGA was associated with a better event-free survival (Figure 4g), suggesting that the GAC/KGA ratio may play a role in cancer progression. However, introduction of either GAC or KGA in BE2C cells or SKNAS cells promoted colony formation (Figure 4h, i), indicating that enhanced glutaminolysis by either GAC or KGA overexpression is pro-proliferative by gaining more ATPs.
+
+
+
JMJD6 regulates alternative splicing of glutaminolysis gene, GLS
+
a. Sashimi plot showing the alternative splicing of GLS after JMJD6 knockdown in BE2C and SK-N-AS cells in duplicates. The number indicates the RNA-seq read counts of exon junction.
+
b. Real time PCR assessing the relative expression of GAC and KGA isoforms after JMJD6 knockdown in triplicates.
+
c. Western blot showing the expression of GAC and KGA isoforms in SK-N-AS, BE2C, SIMA after JMJD6 knockdown for 72 hours.
+
d. KGA and GAC specific reporter analysis showing only KGA-driven luciferase activity is significantly upregulated by JMJD6 knockdown.
+
e. RNA-immunoprecipitation showing JMJD6 interaction with GLS RNA. Top panel shows the western blot analysis of Flag tagged JMJD6 in input, immunoprecipitation (IP) and flowthrough (FT)_fractions. Bottom panel shows RT PCR analysis of enrichment of GAC/KGA bound by JMJD6 in IP and FT fractions.
+
f. Spearman correlation analysis of JMJD6 and GAC/KGA expression levels in two neuroblastoma cohorts GSE45547(left) and GSE120572 (right).
+
g. Kaplan-Meier curve showing the association of GAC and KGA expression levels with event-free survival in a cohort of neuroblastoma (GSE45547).
+
h. Western blotting analysis of expression of KGA and GAC in BE2C cells with indicated antibodies.
+
i. Colony formation assay for BE2C cells overexpressing KGA and GAC for 7 days (left =crystal violet staining, right = quantification of cell density). **p<0.001,
+
j. Western blotting analysis of expression of KGA and GAC in SK-N-AS cells with indicated antibodies.
+
k. Colony formation assay for SK-N-AS cells overexpressing KGA and GAC for 7 days (left =crystal violet staining, right = quantification of cell density). **p<0.001,
+
+
+
+
+
+JMJD6 forms an interaction network with proteins involved in splicing and protein synthesis
+
To understand the mechanism of JMJD6 in regulating splicing in neuroblastoma, we performed an unbiased identification of JMJD6 interacting partners by introducing a flag-tagged JMJD6 into SK-N-AS and BE2C cells, followed by immunoprecipitation to pull down the JMJD6 associated complex and protein identification with mass spectrometry (Figure 5a, Supplementary Table 7). We found that JMJD6 mainly interacted with two classes of proteins which are involved in RNA splicing and protein synthesis in both cell lines, respectively (Figure 5a, Supplementary Figure 6). We then validated the interactions of JMJD6 with splicing factors using immunoprecipitation and western blot, and demonstrated JMJD6 formed a complex with these RNA binding proteins, including RBM39 (Figure 5b), a therapeutic target of high-risk neuroblastoma27. Since JMJD6 also interacted with several molecules involved in protein translation, we investigated if JMJD6 also regulates protein synthesis by using an approach, Click-IT AHA to label the newly synthesized proteins, followed by western blot assessment (Figure 5c). Interestingly, overexpression of JMJD6 greatly reduced total protein synthesis (Figure 5c), suggesting that JMJD6 may antagonize protein production.
+
+
+
JMJD6 forms an interaction network consists of proteins involved in splicing and protein synthesis
+
a. Flag tagged JMJD6 transduced into SK-N-AS cells for immunoprecipitation with anti-Flag followed by protein identification by mass spectrometry. The interacting protein partners of JMJD6 are analyzed by STRING protein network.
+
b. Immunoprecipitation followed by western blot to validate the JMJD6 interacting partners in SK-N-AS cells. IP= immunoprecipitation, FT= flowthrough.
+
c. Click-IT AHA labeling showing the newly synthesized proteins after overexpression of JMJD6 in SK-N-AS cells.
+
d,e. Western blot showing the expression of GAC and KGA isoforms in BE2C (d) SK-N-AS (e), after U2AF2 and CPSF6 knockdown for 72hours.
+
+
+
+
Then we determined if the splicing factors interacting with JMJD6 also regulate GLS isoform expression. Among the splicing factors with which JMJD6 interacted, CPSF6 has been previously shown to regulate the alternative splicing of GLS78. We validated the function of CPSF6 in neuroblastoma cells and found that, indeed, loss of CPSF6 led to a dramatic switch from GAC to KGA in BE2C and SK-N-AS cells (Figure 5d, 5e). Previous studies showed that JMJD6 and U2AF2 (U2AF65) interact to regulate splicing50,79. We also found that knockdown of U2AF2 resulted in a similar phenotype to JMJD6 knockdown in that the expression of KGA isoform was greatly increased in both cell lines (Figure 5d, 5e). These data collectively support the functions of JMJD6 in regulating the splicing of metabolic genes and protein homeostasis in MYC-driven neuroblastoma.
+
+
+JMJD6 regulates production of TCA intermediates and nucleoside triphosphate
+
To further dissect the biological consequences of loss of function of JMJD6, we created stable JMJD6 knockout clones (Supplementary Figure 7) and defined the metabolite spectrum affected by loss of function by using liquid chromatography with tandem mass spectrometry (LC-MS/MS). JMJD6 knockout greatly reduced the production of tricarboxylic acid cycle (TCA) intermediates (i.e., oxoglutarate, fumarate) and nucleoside triphosphate (ATP, CTP, GTP) (Figure 6a), indicating that JMJD6 is a key bioenergetics regulator in cancer cells. Pathway analysis revealed that the reduced metabolites were involved in the Warburg effect, TCA, pentose phosphate pathway, and mitochondrial electron transport chain (Figure 6b), all of which are critical for providing cancer cell bioenergetics for proliferation and survival. Oxoglutarate (α-ketoglutarate) and fumarate are downstream products of glutaminolysis (Figure 6c). We reasoned that cellular metabolites such as glutamate and oxoglutarate may predict the cytotoxic effect of loss-of-function of JMJD6. If cells have higher levels of glutamate and oxoglutarate, they might be less dependent on JMJD6 due to their higher capacity of buffering against reduced glutaminolysis. To test this hypothesis, we used DepMap data that included 225 metabolites in 928 cancer cell lines from over 20 lineages80, and analyzed the correlation of each metabolite with JMJD6 gene dependency. The data showed that cells with high levels of AMP, glutamate, alanine, 2-hydroxyglutarate and 2-oxoglutarate were more resistant to JMJD6 knockout, while cells with high levels of lactose and sucrose were more sensitive to JMJD6 knockout (Figure 6d, Supplementary Table 8). High levels of AMP activate AMP kinase (AMPK), consequently leading to enhanced fatty acid oxidation to stimulate ATP production while alanine can be converted to pyruvate to provide acetyl-CoA to fuel the TCA cycle (Figure 6c). Therefore, high levels of AMP and alanine may provide cells alternative bioenergetics sources for survival. These data further indicate that JMDJ6 function is wired into the regulation of mitochondrial metabolism.
+
+
+
JMJD6 regulates production of citric acid cycle intermediates and NTP
+
a. Heatmap showing the metabolites differentially expressed in SK-N-AS cells (n=5) after JMJD6 knockout (n=5) based on LC-MS/MS analysis.
+
b. Pathway analysis of metabolites downregulated by JMJD6 knockout.
+
c. Pathway cartoon showing the connections of TCA, glycolysis, glutaminolysis, and β-oxidation.
+
d. Correlation of metabolite abundance with JMJD6 dependency. The positive correlation indicates that the higher the abundance of metabolites, the more resistance of cells to JMJD6 knockout. On the contrary, the negative correlation indicates the higher the abundance of metabolites, the more sensitive of cells to JMJD6 knockout.
+
+
+
+
+
+JMJD6 determines the efficacy of indisulam, a molecular glue degrading splicing factor RBM39
+
Dysregulated splicing as a vulnerability of MYC-driven cancers provides a rationale to target neuroblastoma by using splicing inhibitors as a therapeutic approach. We and others have recently reported that indisulam, a “molecular glue” that selectively degrades the splicing factor RBM39, is exceptionally effective at causing tumor regression in multiple high-risk neuroblastoma models without overt toxicity27,28, suggesting indisulam has translational potential. Understanding the factors determining the efficacy of indisulam or any other drug is critical for developing precision therapy, combination therapy or preventing therapy resistance. In addition to complexing together (Figure 5a, b), JMJD6 and RBM39 exhibit significant correlation of co-dependency in cancer cells, namely, cancer cells have similar dependency on JMJD6 and RBM39 for survival (Supplementary Table 3, Figure 7a). These data indicate that JMJD6 may play a role in modulating the effect of indisulam. To test this hypothesis, we performed GSEA to identify the differential pathways between indisulam-sensitive and indisulam-less sensitive neuroblastoma cells. It turned out that histone lysine demethylase (HDM) genes including JMJD6 are the in the only gene signature that is significantly enriched in indisulam-sensitive cells (Figure 7b, c, d). Indeed, knockout of JMJD6 led to partial but significant resistance to indisulam treatment of SK-N-AS cells (Figure 7e-g) and BE2C cells (Figure 7h-j), supporting that cells with high JMJD6 expression are more dependent on RBM39. Since we found that JMJD6 plays a key role in modulating glutaminolysis, we tested if expression of GAC or KGA could affect the activity of indisulam. Interestingly, overexpression of either GAC and KGA renders BE2C and SK-N-AS cells more resistant to indisulam treatment (Figure 7k-n), suggesting that enhanced glutaminolysis may confer therapeutic resistance to spliceosome inhibition.
+
+
+
JMJD6-GAC pathway regulates the response of neuroblastoma cells to indisulam treatment
+
a. Spearman correlation of effects of JMJD6 knockout and RBM39 knockout demonstrating the co-dependency of JMJD6 and RBM39 from DepMAP CRISPR screening data (n=1086). Each dot represents one cell line.
+
b. GSEA analysis for indisulam sensitive vs resistant neuroblastoma cell lines based on CTD2 (Cancer Target Discovery and Development) data showing histone lysine demethylase gene signature is the one that is significantly associated with indisulam response.
+
c. Heatmap from GSEA (b) showing the individual genes in indisulam sensitive and resistant cells.
+
d. JMJD6 expression in indisulam sensitive and resistant neuroblastoma cells. p value calculated by student t test.
+
e. Western blot showing JMJD6 knockout in SK-N-AS cells using indicated antibodies.
+
f. Colony formation of SK-N-AS cells in triplicates with or without JMJD6 knockout treated with different concentrations of indisulam for 7 days, stained with crystal violet.
+
g. Quantification of cell density by using Image J from f. ns=not significant. ** p<0.001, ***p<0.0001
+
h. Western blot showing JMJD6 knockout in BE2C cells using indicated antibodies.
+
i. Colony formation of BE2C cells in triplicates with or without JMJD6 knockout treated with 100nM of indisulam for 5 days, stained with crystal violet.
+
j. Quantification of cell density by using Image J from f. ns=not significant. ** p<0.001, ***p<0.0001
+
k. Colony formation of BE2C cells in triplicates with KGA and GAC overexpression treated with 250nM of indisulam for 5 days, stained with crystal violet.
+
l. Colony formation of SK-N-AS cells in triplicates with KGA and GAC overexpression treated with 100nM of indisulam for 7 days, stained with crystal violet.
+
m. Quantification of cell density by using Image k from k. * p<0.01, **p<0.001
+
n. Quantification of cell density by using Image l from k. **p<0.001
+
+
+
+
+
+
+Discussion
+
MYC is an oncogenic driver of many types of cancer and plays a pivotal role in regulating glycolysis, glutaminolysis, nucleotide and lipid synthesis, and ribosome and mitochondrial biogenesis6. Recent studies have revealed that there is also an interplay between MYC and pre-mRNA splicing machinery7-10,81-84. Pre-mRNA splicing is an essential biological process catalyzed by the spliceosome to produce mature mRNAs85,86. Over 90% of multiexon genes in the human genome undergo alternative splicing87,88, which significantly expands the diversity of the proteome and consequently impacts various biological functions89,90. In this study, we found that the chromosome 17q locus, which is frequently gained in MYC-driven cancers, houses numerous genes essential to cancer cell survival that are implicated in pre-mRNA splicing, ribosome and mitochondrial biogenesis and other biological functions. Particularly, JMJD6, which is located on 17q25 and is amplified in a number of cancer types, physically interacts with a subset of splicing factors such as RBM39 and regulates the alternative splicing of metabolic genes. Depletion of JMJD6 inhibits cancer cell proliferation and impedes tumor growth while overexpression of JMJD6 promotes MYC-mediated tumorigenesis, suggesting that JMJD6 and potentially other 17q genes have oncogenic functions in cellular transformation. Previous studies suggest that JMJD6 and BRD4 interact to regulate gene transcription56,66. However, our unbiased identification of the JMJD6 interactome only identified a subset of proteins involved in mRNA splicing and protein translation in neuroblastoma cells, suggesting that JMD6 may predominantly regulate protein homeostasis to facilitate MYC-mediated transformation. Nevertheless, we cannot exclude the possibility that our immunoprecipitation conditions were too harsh, leading to dissociation of proteins that loosely or dynamically bind to JMJD6.
+
MYC-induced metabolic reprogramming triggers cellular dependency on exogenous glutamine as a source of carbon for mitochondrial membrane potential maintenance and macromolecular synthesis2, leading to “glutamine addiction”2. Glutaminolysis is a process by which GLS and GLS2 convert glutamine to glutamate, which is, in turn, converted by glutamate dehydrogenase or transaminase to 2-oxoglutarate that is further catabolized in the TCA cycle. Additionally, glutamate is a substrate for production of glutathione, an important antioxidant. We previously showed that neuroblastoma relies on MYCN-induced glutaminolysis for survival20. In this study, our RNA-seq barely detected the expression of GLS2 in the neuroblastoma cell models we used, indicating that GLS is the major enzyme that catalyzes glutamine in these model systems. GLS has two isoforms, GAC and KGA, resulting from alternative splicing. KGA is mainly localized in the cytoplasm while GAC is localized in mitochondria and has a higher basal activity73. GAC mRNA levels strongly correlate with the conversion of glutamine to glutamate, as a proxy for GAC activity91. The positive correlation of JMJD6 and GAC suggests that the JMJD6-high tumors have enhanced glutaminolysis activity. CSPF6 and the noncoding RNA CCAT2 have been reported to regulate the splicing of GLS isoforms78,92. Interestingly, we found that depletion of JMJD6 leads to a GLS isoform switch from GAC to KGA, indicating that JMJD6 is involved in alternative splicing of GLS. We further found that JMJD6 physically interacts with CPSF6 in a splicing network and validated that loss of CPSF6 results in remarkable induction of the KGA isoform. We further validated that other splicing factors such as U2AF2 also interact with JMJD6 to regulate the GLS isoform switch. These data indicate that cancer cells can adjust metabolism through alternative splicing to produce enzymes with distinct subcellular localization and activity that promote cellular transformation or progression of an oncogenic phenotype. The cooperation of JMJD6 and MYC in cellular transformation further supports the hypothesis that JMJD6 is needed for metabolic reprogramming triggered by MYC. However, overexpression of either GAC or KGA promotes cell proliferation, suggesting that the switching of KGA/GAC is a cellular fitness mechanism in response to interruption of the spliceosome by adjusting the metabolic rate. Within the tumor microenvironment (i.e., replete and deplete oxygen and nutrient supply) GLS activity is possibly finely tuned through splicing mechanism for adaption.
+
Additionally, we found that JMJD6 physically interacts with a subset of ribosomal proteins that are responsible for protein translation. Interestingly, overexpression of JMJD6 reduces global protein synthesis. A recent study showed that MYC overactivation leads to proteotoxic stress in cells by enhancing global protein synthesis, consequently causing cell death93. The increased global protein synthesis by MYC needs to be buffered through loss of DDX3X, a regulator of ribosome biogenesis and global protein synthesis, for lymphomagenesis93. Our findings suggest that, besides the functions in regulating alternative splicing for metabolism, JMJD6 is involved in MYC-mediated cell transformation by buffering unwanted proteotoxic stress due to high rate of protein synthesis induced by MYC (Figure 8).
+
+
+
Working mechanism of JMJD6 in MYC-driven neuroblastoma.
+
Overactive MYC drives high-load of gene transcription, enhanced protein synthesis and high rate of metabolism, leading to detrimental cellular stresses and consequent cell death (Model a). However, when 17q is amplified, high levels of JMJD6 and other proteins encoded by 17q genes physically interacts with the splicing and translational machineries, enhancing pre-mRNA splicing of metabolic genes such as GLS and inhibiting global protein synthesis, respectively, leading to reduced detrimental stresses and enhanced cancer cell survival and tumorigenesis (Model b). The high levels of JMJD6 predicts high dependency of RBM39, which are more sensitive to indisulam treatment.
+
+
+
+
Neuroblastoma is responsible for as much as 15% of childhood cancer mortality94. With current intensive multimodal therapies, 5-year survival rates for high-risk patients remain less than 50%95-98. In addition, survivors of high-risk disease have a significant risk of developing long-term side effects including subsequent malignant neoplasms due to cytotoxic chemotherapy and radiotherapy99,100. Unfortunately, developing effective precision therapies against high-risk neuroblastoma has been challenging due to the lack of targetable recurrent mutations in neuroblastoma29,30,101. Our previous study showed that indisulam, the splicing inhibitor that targets RBM39, a JMJD6-interacting partner, induced a durable complete response in multiple high-risk neuroblastoma models, supporting its potential use in future clinical trials. Our current study showed that JMJD6 expression is positively correlated with the effect of indisulam, and knockout of JMJD6 confers resistance to indisulam treatment. In line with the biological functions of JMDJ6 in regulating GLS isoform expression and mitochondrial metabolism, overexpression of GAC or KGA also caused resistance to indisulam treatment. These data indicate that JMJD6 could serve as a biomarker that predicts response to indisulam or other splicing inhibitors.
+
+
+Materials and Methods
+
+Cell lines
+
KELLY, SIMA, BE2C, IMR32, SK-N-AS, CHLA20 were cultured in 1X RPMI1640 (Corning, 15-040-CV) supplemented with 10% Fetal Bovine Serum (Sigma-Aldrich, F2442), 1% L-Glutamine (Corning, A2916801). NIH3T3 and 293T cells were cultured in 1X DMEM supplemented with 10% Fetal Bovine Serum (Sigma-Aldrich, F2442), 1% L-Glutamine (Corning, A2916801). All cells were maintained at 37 °C in an atmosphere of 5% CO2. JoMa1 cells were kindly provided by Dr. Schulte (Department of Pediatric Oncology and Hematology, University Children’s Hospital Essen, Essen, Germany) were cultured in NCC Medium: DMEM (4.5 mg/ml Glucose, L-Glutamine, Pyruvate): Ham’s F12 (1:1) was supplemented with: 1% N2-Supplement (Invitrogen, no. 17502-048), 2% B27-Supplement (Invitrogen, no. 17504-044), 10ng/ml EGF (Invitrogen), 1ng/ml FGF (Invitrogen), 100U/ml Penicillin–Streptomycin (Invitrogen) and 10% Chick-Embryo-Extract (Gemini Bio-Products, CA). Neural crest culture medium was supplemented with 200 nM 4-OH-tamoxifen (Sigma no. H7904) in routine culture to ensure nuclear localization of c-MycERT and JoMa1 cell proliferation. JoMa1 cells were grown on cell culture flask/dish coated with fibronectin, NCC-medium supplemented with 200nm 4-OHT was changed daily. Cells were passaged after 3–4 days in culture when 70% confluence was reached (4 X 106 cells/10 cm dish).
+
All human-derived cell lines were validated by short tandem repeat (STR) profiling using PowerPlex® 16 HS System (Promega) once a month. Additionally, a polymerase chain reaction (PCR)-based method was used to screen for mycoplasma once a month employing the LookOut® Mycoplasma PCR Detection Kit (MP0035, Sigma-Aldrich) and JumpStart™ Taq DNA Polymerase (D9307, Sigma-Aldrich) to ensure cells were free of mycoplasma contamination.
+
+
+Retroviral plasmids and retrovirus packaging
+
MSCV-IRES-GFP and MSCV-IRES-mCherry were obtained from St Jude Vector Core. Human JMJD6 and murine MYCN were subcloned into MSCV-IRES-GFP and MSCV-IRES-mCherry, respectively. The MSCV-CMV-CMV-Flag-HA-JMJD6 was purchased from Addgene (Addgene # plasmid 31358). The retrovirus packaging was done as described in the following procedure. Briefly, HEK93T cells were transfected with viral vectors by combining 5μg of target vector, 4.4μg of pMD-old-gag-pol, and 0.6μg of VSV-G plasmids in 400uL of DMEM without serum or L-glutamine. PEIpro transfection reagent (Polyplus 115-010) was added at 2:1 (PEIpro μL: μg of plasmid) per 100mm dish of cells and mixed well, and incubated at RT for at least 20 minutes, prior to adding cells. The following day, fresh medium was added to cells. For 3-4 days, viral media was harvested and replaced twice per day. Viral media was centrifuged at 1500RPM for 10 minutes and filtered through a 0.45um vacuum filter. Virus was concentrated by ultracentrifugation at 28.5kRPM for 2 hours at 4C, aspirated, and resuspended in either OptiMEM or PBS, aliquoted, and frozen at -80C until use. Wasie, add information for GAC/KGA here.
+
+
+siRNA transfection
+
25uM of each siRNA oligo was resuspended in 500uL of prewarmed Opti-MEM, reduced serum medium (Gibco Life technologies #31985-070) in 6 well plates. To each well, 7μL of RNAiMax (Invitrogen Lipofectamine RNAiMAX transfection reagent 13778100) was added, mixed, and left at room temperature for 10 minutes. After incubation, 100,000 cells of each indicated cell line were added to each well in a total of 2mL volume with RPMI medium supplemented with 10% FBS. JMJD6 siRNA#43, 5-CCAAAGUUAUCAAGGAAA-3; JMJD6 siRNA#45, 5-CAGUGAAGAUGAAGAUGAA-3. U2AF2 siRNA#1 AGAAGAAGAAGGUCCGU; U2AF2 siRNA#2 GUGGCAGUUUCAUAUUUG. CPSF6 siRNA#1 GGAUCACCUUCCAAGACA. CPSF6 siRNA#2 AGAACCGUCAUGACGAUU.
+
+
+SDS-PAGE and Western blot
+
Cells were washed twice with ice-cold phosphate-buffered saline (PBS) and directly lysed on ice with 2X sample loading buffer (0.1 M Tris HCl [pH 6.8], 200 mM dithiothreitol [DTT], 0.01% bromophenol blue, 4% sodium dodecyl sulfate [SDS] and 20% glycerol). On ice, cell lysates were sonicated once with a 5 second bursts at 40% amplitude output (Sonics, VIBRA CELL) followed by 25 minutes heating at 95 °C. After the cell lysates were centrifuged at 13,000 × g at room temperature for 2 mins, 10-20 µl of the cell lysates were separated on 4-15% Mini-PROTEAN® TGX™ Stain-FreeTM Protein Gels from Bio-Rad and transferred to methanol-soaked polyvinylidene difluoride (PVDF) membranes (Millipore). Lysates for RBM39 G268V mutant cell lines and DCAF15 genetically modified cells were generated as previously described102. Membranes were blocked in PBS buffer supplemented with 0.1% TWEEN 20 and 5% skim milk (PBS-T) and incubated for 1 hour at room temperature under gentle horizontal shaking. Membranes were incubated overnight at 4 °C with the primary antibodies. The next day, membranes were washed 3 times (for 5 minutes) with PBS-T at room temperature. Protected from light, membranes were then incubated with goat anti-mouse or goat anti-rabbit HRP-conjugated secondary antibodies (1:5,000) for 1 hour at room temperature, followed by three 5-minite washes with PBS-T at room temperature. Lastly, membranes were incubated for 1 minute at room temperature with SuperSignal West Pico PLUS Chemiluminescent Substrate (34580, Thermo Fisher Scientific) and the bound antigen-antibody complexes were visualized using Odyssey Fc Imaging System (LI-COR Corp., Lincoln, NE).
+
+
+RNA extraction and RT-PCR isoforms of GLS
+
RNA was extracted using RNeasy® Plus Mini Kit (Qiagen, reference # 74136) following the manufacturer’s protocol. cDNA was prepared in 20ul reaction from 500ng of total RNA using Superscript™ IV First Strand Synthesis System (Invitrogen, reference # 1809105) kit. Real-time PCR reactions were run in triplicates (n=3) in the 7500 Real-time PCR system by Applied Biosystems (Thermo Fisher Scientific) using power SYBR Green PCR master mix (Applied Biosystems, reference # 4367660). ΔΔCT methods were applied to analyze the results. The following primers were used to perform the quantitative Real-time PCR-GAPDH (Forward: AACGGGAAGCTTGTCATCAATGGAAA, Reverse: GCATCAGCAGAGGGGGCAGAG), GAC (Forward: GAGGTGCTGGCCAAAAAGCCT, Reverse: AGGCATTCGGTTGCCCAAACT), KGA (Forward: CTGCAGAGGGTCATGTTGAA, Reverse: ATCCATGGGAGTGTTATTCCA).
+
+
+Lentiviral packaging of pLenti and shRNA
+
The GAC and KGA cDNAs were synthesized by Genscript company and cloned into pLenti vector. The TRC lentiviral-based shRNA knockdown plasmids for JMJD6 were purchased from Horizon Discovery (sh#46: RHS3979-201781036, TTAAACCAGGTAATAGCTTCG; sh#47: RHS3979-201781037, ATCTTCACTGAGTAGCCATCG) The lentiviral shJMJD6 and shControl (pLKO.1) particles were packaged by Vector Lab at St Jude. Briefly, HEK293T cells were transfected with shRNA constructs and helper plasmids (pCAG-kGP1-1R, pCAG4-RTR2, and pHDM-G). The 48- and 72-hr post-transfection replication-incompetent lentiviral particles were harvested and transduced into cells with 8μg/ml of polybrene. 48 hours later, 1 μg/ml of puromycin was added for selection for additional 48 hours before injection into mice or immunoblotting.
+
+
+JMJD6 CRISPR KO method
+
Genetically modified neuroblastoma cells were generated by using CRISPR-Cas9 technology. Briefly, 400,000 NB cells were transiently co-transfected with 100pmol of chemically modified gRNA (GGACTCTGGAGCGCCTAAAA) (Synthego), 33pmol of Cas9 protein (St. Jude Protein Production Core), 200ng of pMaxGFP (Lonza), and, using solution P3 and program DS-150 in small cuvettes according to the manufacturer’s recommended protocol. Five days post-nucleofection, cells were sorted for GFP+ (transfected) cells and plated as single cells into 96-well plates. Cells were clonally expanded and screened for the desired modification using targeted next generation sequencing followed by analysis with CRIS.py (https://pubmed.ncbi.nlm.nih.gov/30862905/).
+
+
+Colony formation for JoMa1 cells
+
Matrigel was kept at 4°C to being liquified for 6 hours. 50μL of Matrigel per 1 square centimeter area was added to 24-well plate without air bubble. The 24-well plate was kept at 37°C in cell culture incubator till it was solidified. 200 of JoMa1 cells transduced with GFP, JMJD6, MYCN, MYCN+JMJD6 in DMEM:F12 enriched media without tamoxifen were seeded onto the 24-well coated with Matrigel. This was done in triplicate. Cells were checked daily, and media were changed every 3 days without disturbing the Matrigel by removing and adding media gently. To stain the colonies, cells were fixed by formaldehyde (3.7% in PBS) for 2 min at room temperature, followed by permeabilization with 100% methanol (not ice-cold) for 20min at room temperature. The colonies were stained by 0.4% crystal violet.
+
+
+Crystal Violet Staining
+
After removing media, cells were washed with Dulbecco’s phosphate buffered saline without calcium or magnesium (DPBS, Lonza) and treated with 4% formaldehyde in PBS (PFA) for 20 minutes. Once PFA was removed, cells were stained with 0.1% crystal violet stain for 1 hour. KGA/GAC overexpression colony formation: 5000 cells were plated of BE2C control, KGA and GAC overexpressing cells and were culture for 7 days; 10,000 cells were plated of SKNAS control, KGA and GAC overexpressing cells and were culture for 7 days (n=3). After 7 days, medium was removed and cells were washed with Dulbecco’s PBS (DPBS) (DPBS, Lonza) and treated with 4% formaldehyde in PBS [paraformaldehyde (PFA)] for 30 min. PFA was later removed and cells were stained with 0.1% crystal violet stain for 1 hour. Experiments were repeated twice. Indisulam treatment on WT and JMJD6-KO cell lines: JMJD6-WT and KO cells were plated in 12-well plate (50,000 cells/well for SKNAS) and 6-well plate (5,000 cells/well for BE2C) (n=3). Next day, cells were treated with Indisulam with indicated concentration for 7 days (SKNAS cells) and 5 days (BE2C cells). Crystal violet staining was performed to visualize and quantify the colony formation. Experiments were repeated twice. Indisulam treatment on KGA/GAC overexpressing cell lines: 10,000 BE2C and 100,000 SKNAS control, KGA and GAC overexpressing cells were plated in 6-well plate (n=3). Next day, cells were treated with Indisulam for 7 days. Crystal violet staining was performed to visualize and quantify the colony formation. Experiments were repeated twice.
+
+
+Click-iT AHA labeling assay for metabolic labeling of newly synthesized proteins
+
Click-iT was performed as previously described. Briefly, cells were plated at 5 million cells per 100mm dish in RPMI supplemented with 10% FBS. Cells were washed with warm PBS and replaced with methionine-free medium (Thermo 21013024 supplement with glutamine and sodium pyruvate) for 1 hour at 37° C in 5% CO2. Following, fresh methionine-free media containing 50μM of Click-iT AHA (L-azidohomoalanine) (Thermo C10102) was added to the cells for 2 hours at 37°C. After AHA-labeling, cells were washed with warm PBS and lysed with 1% SDS, 50mM Tris-HCl, (pH 8.0) supplemented with phosphatase inhibitors (PhosSTOP, Sigma) and protease inhibitors (cOmplete mini, Roche) by applying the buffer directly to the plate, incubating the cells on ice for 30 minutes, tilting the plates, and collecting the lysate. Lysates were briefly sonicated, vortexed for 5 minutes, and centrifuged at 18,000xg for 5 minutes at 4°C. Total protein quantification was assayed using the EZQ Protein Quantification Kit (Thermo R33200) according to the manufacturer’s protocol and results were read on a fluorescence-based microplate reader (BioTek Synergy 2). Click chemistry of the biotin-alkyne (PEG4 carboxamide-propargyl biotin) (Thermo B10185) to the AHA-labeled lysates was performed using the Click-iT Protein Reaction Buffer Kit (Thermo C10276) using a concentration of 40μM biotin-alkyne per click reaction (and no biotin-alkyne added for controls). Following the click reaction, samples were either assayed for total biotinylated protein by following the manufacturer’s protocol. For total biotinylated protein, briefly, 600μL of methanol, 150μL of chloroform, and 400μL of megaOhm water was sequentially added and vortexed, followed by centrifugation at 18,000xg for 5 minutes. The upper aqueous phase was discarded, and 450μL of methanol was added, vortexed, and centrifuged again at 18,000xg for 5 minutes. This methanol step was performed in duplicate to remove residual reaction components. Protein pellets were allowed to air dry and resuspended in a suitable volume of sample buffer and heated prior to western blot analysis.
+
+
+Immunoprecipitation
+
5X106 BE2C and SK-N-AS cells expressing Flag-JMJD6 were cultured in 150cm dish with complete RPMI media. Cells were washed twice with cold PBS after reaching 95% confluency, then lysed in 1ml lysis buffer (50mM Tris-HCl, pH 7.4, 150mM NaCl, 1mM EDTA, 1% Trion-X100 with complete protease inhibitors (Sigma 11836170001, added fresh) and PhosSTOP (Sigma 4906845001). Cells were scrapped into a 1.5mL Eppendorf tube and incubated on ice for 15min, which were mixed by vortex every 5 min. Cell lysates were spun by 135000rpm for 10min at 4C. The supernatant was transferred to a new tube. The cell lysates were subject to immunoprecipitation using M2 anti-Flag beads (Sigma, M8823) overnight by rocking at 4°C. The following day, beads were washed 3X with buffer and eluted with 5 packed gel volumes of FLAG peptide in TBS buffer (3uL of stock FLAG peptide at 5ug/uL per 100uL of TBS buffer) while rotating at 4°C for 30 minutes. Beads were briefly spun and the supernatant was removed from the beads (eluate). This elution step was repeated one more time and pooled with the first eluate. Prior to western blot, input, flow throughs, and elution samples were processed by adding 4X sample buffer supplemented with 50mM DTT and heated at 75°C for 10 minutes prior to running on a gel.
+
+
+RNA-immunoprecipitation
+
SK-N-AS cells expressing Flag-JMJD6 were grown in a 10-cm dish in RPMI complete media. After 70% confluency, cells were washed with cold PBS twice and then were subject to lysis with Polysome Lysis Buffer (100mM KCl, 5mM MgCl2, 10mM HEPES, pH7.0, 0.5% NP-40, 1mM DTT, 100 U/ml RAasin RNase inhibitor (Promega, N2511), 2mM vanadyl ribonucleoside complexes solution (Sigma, 94742), 25μL protease inhibitor cocktail for mammalian cells (Sigma, P8340)). Cell lysates were precleared with magnetic IgG beads for 1 hour. The cell lysates were subject to immunoprecipitation using M2 anti-Flag beads (Sigma, M8823) overnight by rocking at 4°C. The same amounts of lysates were saved at -80°C for input RNA extraction. The beads were washed with 250 μL Polysome Lysis Buffer for 4 times, flowed by washing with Polysome Lysis Buffer containing 1M urea. RNA was released by adding 150μL of Polysome Lysis Buffer containing 0.1% SDS and 45μg protease K (Ambion, AM2548) and incubated at 50°C for 30min. RNA was extracted with phenol-chloroform-isoamyl alcohol mixture (Sigma, 77618). RNA was recovered by adding 2μL of GlycoBlue (15mg/ml, Ambion, AM9516), 36μL of 3M sodium acetate and 750μL ethanol followed by incubation at -20°C for overnight. RNA was precipitated with 70% ethanol and air dried, followed by resuspension with RNase-free water followed by DNaseI (Promega, M6101) treatment to remove genomic DNA. The resultant RNAs were subjected to RT-qPCR analysis using 3 sets of GAC and KGA primers and 18S rRNA as control. 18S rRNA F: GCTTAATTTGACTCAACACGGGA; 18S rRNA R: AGCTATCAATCTGTCAATCCTGTC. GLS-GACiso_F: GAGGTGCTGGCCAAAAAGCCT; GLS-GACiso_R: AGGCATTCGGTTGCCCAAACT. GLS-KGAiso_F: CTGCAGAGGGTCATGTTGAA; GLS-KGAiso_R: ATCCATGGGAGTGTTATTCCA. KGA_set2_F: GCAGCCTCCAGGTGCTTTCA; KGA_set2_R: GTAATGGGAGGGCAGTGGCA. KGA_set3_F: TGCCCGACACTGCCCTTTAG; KGA_set3_R: CCTGCCAGACAGACAACAGCA. GAC_set2_F: TGCTTCTCAAGGCCTTACTGC; GAC_set2_R: AGGCATTCGGTTGCCCAAACT. GAC_set3_F: CCTTCTAGAGGTGCTGGCCAAA; GAC_set3_R: TGCAACACAAATATGCAGTAAGGC. For validation of protein immunoprecipitation, 20% of beads after overnight incubation were removed and processed as follows: Beads were washed 3X with buffer and eluted with 5 packed gel volumes of FLAG peptide in TBS buffer (3μL of stock FLAG peptide at 5μg/μL per 100uL of TBS buffer) while rotating at 4°C for 30 minutes. Beads were briefly spun and the supernatant was removed from the beads (eluate). This elution step was repeated one more time and pooled with the first eluate.
+
+
+Identification of JMJD6 interacting partners by LC-MS/MS
+
Protein samples were run on a short gel as described in a previously published protocol103. Proteins in the gel bands were reduced with dithiothreitol (DTT) (Sigma) and alkylated by iodoacetamide (IAA) (Sigma). The gel bands were then washed, dried, and rehydrated with a buffer containing trypsin (Promega). Samples were digested overnight, acidified and the resulting peptides were extracted. The extracts were dried and reconstituted in 5% formic acid. The peptide samples were loaded on a nanoscale capillary reverse phase C18 column by a HPLC system (Thermo EASY-nLC 1000) and eluted by a gradient. The eluted peptides were ionized and detected by a mass spectrometer (Thermo LTQ Orbitrap Elite). The MS and MS/MS spectra were collected over a 90-min liquid chromatography gradient. Database searches were performed using Sequest (version 28 revision 13) search engine against a composite target / decoy Uniprot human protein database. All matched MS/MS spectra were filtered by mass accuracy and matching scores to reduce protein false discovery rate to <1%. Spectral counts, matching to individual proteins reflect their relative abundance in one sample after the protein size is normalized. The spectral counts between samples for a given protein was used to calculate the p-value based on G-test 104.
+
+
+Metabolome profiling by LC-MS/MS
+
MJD6 knockout or parental SK-N-AS cells were cultured in 6-well plates to ∼85% confluence and washed with 2 mL ice cold 1X Phosphate-Buffered Saline (PBS). The cells were then harvested in 300 µL freezing 80% acetonitrile (v/v) into 1.5 mL tubes and lysed in the presence of 0.5mm Zirconia/silica beads by Bullet Blender (Next Advance) at 4 °C until the sample were homogenized. The resulting lysate was then centrifuged at 21,000 x g for 5 min and the supernatant was dried by speedvac. The samples were resuspended in 50 µL of 1% acetonitrile plus 0.1% trifluoroacetic acid, and separated by Ultra-C18 Micro spin columns (Harvard apparatus) into hydrophilic metabolites (flow through) and hydrophobic metabolites (eluent of 125 µL of 80% acetonitrile plus 0.1% trifluoroacetic acid). Ten µL of hydrophilic metabolites were dried, reconstituted in 3 µL of 66% acetonitrile and analyzed by a ZIC-HILIC column (150 × 2.1 mm, EMD Millipore) coupled with a Q Exactive HF Orbitrap MS (Thermo Fisher) in negative mode and metabolites were eluted within a 45 min gradient (buffer A: 10mM ammonium acetate in 90% acetonitrile (pH=8); buffer B: 10mM ammonium acetate in 100% H2O (pH=8)). Twenty µL of hydrophobic metabolites were dried and resuspend in 3 µL of 5% formic acid followed by separation with a self-packed nanoC18 column (75 μm × 15 cm with 1.9 µm C18 resin from Dr. Maisch GmbH) and detection with a Q Exactive HF Orbitrap MS (Thermo Fisher) in positive mode. Metabolites were eluted within a 50 min gradient (buffer A: 0.2% formic acid in H2O; buffer B: 0.2% formic acid in acetonitrile). MS settings for both types of samples included MS1 scans (120,000 resolution, 100-1000 m/z, 3 x 106 AGC and 50 ms maximal ion time) and 20 data-dependent MS2 scans (30,000 resolution, 2 x 105 AGC, ∼45 ms maximal ion time, HCD, Stepped NCE (50, 100, 150), and 20 s dynamic exclusion). A mix of all samples served as quality control was injected in the beginning, middle and the end of the samples to monitor the signal stability of the instrument. The data analysis was performed by a recently developed software suite JUMPm. Raw files were converted to mzXML format followed by peak feature detection for individual sample and feature alignment across samples. Metabolite identification was supported by matching the retention time, accurate mass/charge (m/z) ratio, and MS/MS fragmentation data to our in-house authentic compound library and the matching of m/z and MS/MS fragmentation data to, downloaded experimental MS/MS library (MoNA, https://mona.fiehnlab.ucdavis.edu/), in-silico database generated from Human Metabolome Database (HMDB), and mzCloud (https://mzcloud.org). Peak intensities were used for metabolite quantification. The data was normalized by both cell numbers (before data collection) and trimmed median intensity of all features across samples (post data collection).
+
+
+Differential gene expression and gene set enrichment analysis (GSEA) for RNA-seq experiments
+
Total RNA from cells and tumor tissues were performed using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Paired-end sequencing was performed using the High-Seq platform with 100bp read length. Total stranded RNA sequencing data were processed by the internal AutoMapper pipeline. Briefly the raw reads were firs trimmed (Trim-Galore version 0.60), mapped to human genome assembly (GRCh38) (STAR v2.7) and then the gene level values were quantified (RSEM v1.31) based on GENCODE annotation (v31). Low count genes were removed from analysis using a CPM cutoff corresponding to a count of 10 reads and only confidently annotated (level 1 and 2 gene annotation) and protein-coding genes are used for differential expression analysis. Normalization factors were generated using the TMM method, counts were then transformed using voom and transformed counts were analyzed using the lmFit and eBayes functions (R limma package version 3.42.2). The significantly up- and down-regulated genes were defined by at least 2-fold changes and adjusted p-value < 0.05. Then gene set enrichment analysis (GSEA) was carried out using gene-level log2 fold changes from differential expression results against gene sets in the Molecular Signatures Database (MSigDB 6.2) (gsea2 version 2.2.3).
+
+
+RNA splicing analysis
+
After mapping RNA-seq data, rMATS v4.1.0 was used for RNA alternative splicing analysis by using the mapped BAM files as input. Specifically, five different kinds of alternative splicing events were identified, i.e., skipped exon (SE), alternative 5’-splicing site (A5SS), alternative 3’-splicing site (A3SS), mutually exclusive exon (MXE) and intron retention (RI). To keep consistent, the same GTF annotation reference file for mapping was used for rMATS. For stranded RNA-seq data, the argument “--libType fr-firststrand” was applied. To process reads with variable lengths, the argument “--variable-read-length” was also used for rMATS. To select statistically significantly differential splicing events, the following thresholds were used: FDR <0.05 and the absolute value of IncLevelDifference > 0.1. For visualization, the IGV Genome Browser was used to show the sashimi plots of splicing events. To investigate the genome-wide correlations of differential splicing between two genotypes (e.g., shRNA knockdown of JMJD6 and non-target shRNA in cells), we extracted splice junctions for all samples of both genotypes of interest from the STAR105 output files suffixed with “SJ.out.tab”, which contain high confidence collapsed splice junctions. Only those unique mapped reads crossing the junctions were considered. By extracting the union of the unique junction positions, we constructed a unified junction-read feature vector for each sample. Then, we normalized the junction-read vectors of each sample with TMM method in “voom” and “limma” and R package, assuming a negative binomial distribution. Next, we averaged the junction-read vectors for samples of the same genotype. The gene level expression was estimated based on the canonical junctions from the most abundant isoforms estimated for each gene. The fold changes of exon junctions significantly deviated from gene level changes were regarded as differentially spliced junctions for between cell-line comparisons.
+
+
+Data mining
+
JMJD6, GAC and KGA expression in tumor tissues were downloaded from R2 (https://portals.broadinstitute.org/ccle), Kocak dataset GSE45547 (649 samples) and Fischer dataset GSE120572 (394 samples). In both datasets, the probe UKv4_A_23_P311616 represented JMJD6, the probe UKv4_A_23_P308800 represented GAC and UKv4_A_23_P39766 represented KGA. JMJD6 expression data from the RNA-seq data of various pediatric cancer tissues were downloaded from St Jude cloud (https://pecan.stjude.cloud/). The copy number alterations of JMJD6 and the related Kaplan-Meier analysis were downloaded from cBioportal (cbioportal.org). The data for correlation of metabolite abundance and JMJD6 knockout effect were downloaded from DepMap (https://depmap.org/portal/).
+
+
+Pathway network analysis
+
The 114 essential fitness genes to neuroblastoma cell survival identified through genome-wide CRISPR/Cas9 library screen were uploaded into STRING program (https://string-db.org) for network interaction analysis with confidence threshold 0.15. The resulting network was then uploaded into Cytoscape program for presentation106. The clusters were grouped based on the biological functions of each gene.
+
Copy number analysis of JMJD6 and other genes encoding JmjC domain histone demethylases from St Jude neuroblastoma cohort.
+
Somatic copy number alternations (SCNA) were determined by CONSERTING (PMID: 25938371) for each pair of tumor and normal samples. The normalize read depth ratio (log2 ratio) for the CNV segments with JmjC-domain containing proteins were extracted and used for CNV heatmap generation (https://CRAN. R-project. org/package= pheatmap) and hierarchical clustering of samples.
+
+
+Xenograft studies
+
+
(1) shRNA-mediated JMJD6 knockdown. Neuroblastoma cells were transduced with shRNA lentiviral particles targeting JMJD6. 48 hours later, 1 μg/ml of puromycin was added for selection for additional 48 hours. Cancer cells (5x106) were mixed with Matrigel (1:1 ratio in volume) and subcutaneously injected into the flank sites of NSG mice. (2) JMJD6 and MYC-mediated transformation. After JoMa1 cells were transduced with GFP, JMJD6, MYCN and JMJD6/MYCN, 104 cells per group were mixed with Matrigel (1:1 ratio in volume) and subcutaneously injected into the flank sites of NSG mice. Mice were sacrificed when they reached the humane endpoint. Tumors were measured by using electronic calipers, and volumes calculated as width ρχ/6 xd3 where d is the mean of two diameters taken at right angles.
+
+
+
+Statistical analysis
+
All quantitative data are presented as mean ± SD. Unpaired Student’s t test was performed for comparison of two groups. Spearman correlation was used to assess the relationship between two variables. Kaplan-Meier method was used to estimate the survival rate. Mann-Whitney rank test (two-sided) was used to compare the tumor volume between two groups at every time point. P-values across multiple time points were adjusted for multiple comparison using the Holm-SidaK method. p<0.05 was considered as statistically significant. All the statistical analyses, except where otherwise noted, were performed using GraphPad Prism (v9).
+
+
+Data accessibility
+
GEO accession number: GSE185867 To review GEO accession GSE185867: Go to https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ncbi.nlm.nih.gov%2Fgeo%2Fquery%2Facc.cgi%3Facc%3DGSE185867&data=04%7C01%7Cjun.yang2%40stjude.org%7C56d8ac619531468173d608d98ea927d8%7C22340fa892264871b677d3b3e377af72%7C0%7C0%7C637697679446209542%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=XtZl1ttoOPBvfhwmX9uCcwVN1Yg02qDsvcsHdJv58xw%3D&reserved=0 Enter token qzixwsoohbsxjij into the box
+
+
+
+
+
+Acknowledgements
+
We thank the staff of the St. Jude Animal Resource Center and Hartwell Center for their dedication and expertise. The work was supported by American Cancer Society-Research Scholar (130421-RSG-17-071-01-TBG, J.Y.), National Cancer Institute (1R01CA229739-01, J.Y., R01CA266600, J.Y.). The work was also supported by the American Lebanese Syrian Associated Charities (ALSAC). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
+
+
+Author contributions
+
C.J., W.Q., S.S., H.H., D.H.B. D.H performed experiments. G.W., H.J., T-C.C., D.F., J-H.C performed proteomic, genomic and RNA-seq analysis. S.M.S and S.M.P-M provided CRISPR knockout. R.W., K.F. provided key reagents. A.M.D, A.M., J.P., and J.Y. supervised the studies. J.Y conceived the project and wrote the manuscript with help from co-authors.
+
+
+References
+Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell144, 646–674 (2011). https://doi.org:10.1016/j.cell.2011.02.013
+Pavlova, N. N. & Thompson, C. B. The Emerging Hallmarks of Cancer Metabolism. Cell Metab23, 27–47 (2016). https://doi.org:10.1016/j.cmet.2015.12.006
+Vazquez, A.et al.Cancer metabolism at a glance. J Cell Sci129, 3367–3373 (2016). https://doi.org:10.1242/jcs.181016
+DeBerardinis, R. J. & Chandel, N. S. Fundamentals of cancer metabolism. Sci Adv2, (2016). https://doi.org:10.1126/sciadv.1600200
+Kalkat, M.et al.MYC Deregulation in Primary Human Cancers. Genes (Basel) 8 (2017). https://doi.org:10.3390/genes8060151
+Stine, Z. E., Walton, Z. E., Altman, B. J., Hsieh, A. L. & Dang, C. V. MYC, Metabolism, and Cancer. Cancer Discov5, 1024–1039 (2015). https://doi.org:10.1158/2159-8290.CD-15-0507
+Hsu, T. Y.et al.The spliceosome is a therapeutic vulnerability in MYC-driven cancer. Nature525, 384–388 (2015). https://doi.org:10.1038/nature14985
+Hirsch, C. L.et al.Myc and SAGA rewire an alternative splicing network during early somatic cell reprogramming. Genes Dev29, 803–816 (2015). https://doi.org:10.1101/gad.255109.114
+Koh, C. M.et al.MYC regulates the core pre-mRNA splicing machinery as an essential step in lymphomagenesis. Nature523, 96–100 (2015). https://doi.org:10.1038/nature14351
+Phillips, J. W.et al.Pathway-guided analysis identifies Myc-dependent alternative pre-mRNA splicing in aggressive prostate cancers. Proc Natl Acad Sci U S A117, 5269–5279 (2020). https://doi.org:10.1073/pnas.1915975117
+Anczukow, O. & Krainer, A. R. The spliceosome, a potential Achilles heel of MYC-driven tumors. Genome Med7, (2015). https://doi.org:10.1186/s13073-015-0234-3
+Zhang, S.et al.MYCN controls an alternative RNA splicing program in high-risk metastatic neuroblastoma. Cancer Lett371, 214–224 (2016). https://doi.org:10.1016/j.canlet.2015.11.045
+Gustafson, W. C. & Weiss, W. A. Myc proteins as therapeutic targets. Oncogene29, 1249–1259 (2010). https://doi.org:10.1038/onc.2009.512
+Weiss, W. A., Aldape, K., Mohapatra, G., Feuerstein, B. G. & Bishop, J. M. Targeted expression of MYCN causes neuroblastoma in transgenic mice. EMBO J16, 2985–2995 (1997). https://doi.org:10.1093/emboj/16.11.2985
+Zhu, S.et al.Activated ALK collaborates with MYCN in neuroblastoma pathogenesis. Cancer Cell21, 362–373 (2012). https://doi.org:10.1016/j.ccr.2012.02.010
+Tao, L.et al.MYCN-driven fatty acid uptake is a metabolic vulnerability in neuroblastoma. Nat Commun13, 3728 (2022). https://doi.org:10.1038/s41467-022-31331-2
+Khan, A.et al.Targeting metabolic activity in high-risk neuroblastoma through Monocarboxylate Transporter 1 (MCT1) inhibition. Oncogene39, 3555–3570 (2020). https://doi.org:10.1038/s41388-020-1235-2
+Gamble, L. D.et al.Inhibition of polyamine synthesis and uptake reduces tumor progression and prolongs survival in mouse models of neuroblastoma. Sci Transl Med11 (2019). https://doi.org:10.1126/scitranslmed.aau1099
+Xia, Y.et al.Metabolic Reprogramming by MYCN Confers Dependence on the Serine-Glycine-One-Carbon Biosynthetic Pathway. Cancer Res79, 3837–3850 (2019). https://doi.org:10.1158/0008-5472.CAN-18-3541
+Wang, T.et al.MYCN drives glutaminolysis in neuroblastoma and confers sensitivity to an ROS augmenting agent. Cell Death Dis9, (2018). https://doi.org:10.1038/s41419-018-0295-5
+Bansal, M., Gupta, A. & Ding, H. F. MYCN and Metabolic Reprogramming in Neuroblastoma. Cancers (Basel) 14 (2022). https://doi.org:10.3390/cancers14174113
+Olsen, T. K.et al.DHODH is an independent prognostic marker and potent therapeutic target in neuroblastoma. JCI Insight7 (2022). https://doi.org:10.1172/jci.insight.153836
+Alborzinia, H.et al.MYCN mediates cysteine addiction and sensitizes neuroblastoma to ferroptosis. Nat Cancer3, 471–485 (2022). https://doi.org:10.1038/s43018-022-00355-4
+Guo, X., Chen, Q. R., Song, Y. K., Wei, J. S. & Khan, J. Exon array analysis reveals neuroblastoma tumors have distinct alternative splicing patterns according to stage and MYCN amplification status. BMC Med Genomics4, (2011). https://doi.org:10.1186/1755-8794-4-35
+Shi, Y.et al.Aberrant splicing in neuroblastoma generates RNA-fusion transcripts and provides vulnerability to spliceosome inhibitors. Nucleic Acids Res (2021). https://doi.org:10.1093/nar/gkab054
+Shi, Y.et al.Aberrant splicing in neuroblastoma generates RNA-fusion transcripts and provides vulnerability to spliceosome inhibitors. Nucleic Acids Res49, 2509–2521 (2021). https://doi.org:10.1093/nar/gkab054
+Singh, S.et al.Targeting the spliceosome through RBM39 degradation results in exceptional responses in high-risk neuroblastoma models. Sci Adv7, eabj5405 (2021). https://doi.org:10.1126/sciadv.abj5405
+Nijhuis, A.et al.Indisulam targets RNA splicing and metabolism to serve as a therapeutic strategy for high-risk neuroblastoma. Nat Commun13, 1380 (2022). https://doi.org:10.1038/s41467-022-28907-3
+Pugh, T. J.et al.The genetic landscape of high-risk neuroblastoma. Nat Genet45, 279–284 (2013). https://doi.org:10.1038/ng.2529
+Molenaar, J. J.et al.Sequencing of neuroblastoma identifies chromothripsis and defects in neuritogenesis genes. Nature483, 589–593 (2012). https://doi.org:10.1038/nature10910
+Hagenbuchner, J., Kiechl-Kohlendorfer, U., Obexer, P. & Ausserlechner, M. J. BIRC5/Survivin as a target for glycolysis inhibition in high-stage neuroblastoma. Oncogene35, 2052–2061 (2016). https://doi.org:10.1038/onc.2015.264
+MacArthur, I. C.et al.Prohibitin promotes de-differentiation and is a potential therapeutic target in neuroblastoma. JCI Insight5 (2019). https://doi.org:10.1172/jci.insight.127130
+Milosevic, J.et al.PPM1D Is a Therapeutic Target in Childhood Neural Tumors. Cancers (Basel) 13 (2021). https://doi.org:10.3390/cancers13236042
+Meitinger, F.et al.TRIM37 controls cancer-specific vulnerability to PLK4 inhibition. Nature585, 440–446 (2020). https://doi.org:10.1038/s41586-020-2710-1
+Garcia-Lopez, J.et al.Large 1p36 Deletions Affecting Arid1a Locus Facilitate Mycn-Driven Oncogenesis in Neuroblastoma. Cell Rep30, 454–464 e455 (2020). https://doi.org:10.1016/j.celrep.2019.12.048
+Henrich, K. O.et al.CAMTA1, a 1p36 tumor suppressor candidate, inhibits growth and activates differentiation programs in neuroblastoma cells. Cancer Res71, 3142–3151 (2011). https://doi.org:10.1158/0008-5472.CAN-10-3014
+Liu, Z.et al.CASZ1, a candidate tumor-suppressor gene, suppresses neuroblastoma tumor growth through reprogramming gene expression. Cell Death Differ18, 1174–1183 (2011). https://doi.org:10.1038/cdd.2010.187
+Laut, A. K.et al.CHD5 inhibits metastasis of neuroblastoma. Oncogene41, 622–633 (2022). https://doi.org:10.1038/s41388-021-02081-0
+Higashi, M.et al.Retinoic acid-induced CHD5 upregulation and neuronal differentiation of neuroblastoma. Mol Cancer14, (2015). https://doi.org:10.1186/s12943-015-0425-y
+Fujita, T.et al.CHD5, a tumor suppressor gene deleted from 1p36.31 in neuroblastomas. J Natl Cancer Inst100, 940–949 (2008). https://doi.org:10.1093/jnci/djn176
+Fell, S. M.et al.Neuroblast differentiation during development and in neuroblastoma requires KIF1Bbeta-mediated transport of TRKA. Genes Dev31, 1036–1053 (2017). https://doi.org:10.1101/gad.297077.117
+Li, S.et al.The 1p36 Tumor Suppressor KIF 1Bbeta Is Required for Calcineurin Activation, Controlling Mitochondrial Fission and Apoptosis. Dev Cell36, 164–178 (2016). https://doi.org:10.1016/j.devcel.2015.12.029
+Chen, Z. X.et al.RNA helicase A is a downstream mediator of KIF1Bbeta tumor-suppressor function in neuroblastoma. Cancer Discov4, 434–451 (2014). https://doi.org:10.1158/2159-8290.CD-13-0362
+Cole, K. A.et al.A functional screen identifies miR-34a as a candidate neuroblastoma tumor suppressor gene. Mol Cancer Res6, 735–742 (2008). https://doi.org:10.1158/1541-7786.MCR-07-2102
+Yu, F.et al.RUNX3 interacts with MYCN and facilitates protein degradation in neuroblastoma. Oncogene33, 2601–2609 (2014). https://doi.org:10.1038/onc.2013.221
+Bown, N.et al.Gain of chromosome arm 17q and adverse outcome in patients with neuroblastoma. N Engl J Med340, 1954–1961 (1999). https://doi.org:10.1056/NEJM199906243402504
+Althoff, K.et al.A Cre-conditional MYCN-driven neuroblastoma mouse model as an improved tool for preclinical studies. Oncogene34, 3357–3368 (2015). https://doi.org:10.1038/onc.2014.269
+Bottger, A., Islam, M. S., Chowdhury, R., Schofield, C. J. & Wolf, A. The oxygenase Jmjd6--a case study in conflicting assignments. Biochem J468, 191–202 (2015). https://doi.org:10.1042/BJ20150278
+Chang, B., Chen, Y., Zhao, Y. & Bruick, R. K. JMJD6 is a histone arginine demethylase. Science318, 444–447 (2007). https://doi.org:10.1126/science.1145801
+Webby, C. J.et al.Jmjd6 catalyses lysyl-hydroxylation of U2AF65, a protein associated with RNA splicing. Science325, 90–93 (2009). https://doi.org:10.1126/science.1175865
+Kwok, J., O’Shea, M., Hume, D. A. & Lengeling, A. Jmjd6, a JmjC Dioxygenase with Many Interaction Partners and Pleiotropic Functions. Front Genet8, (2017). https://doi.org:10.3389/fgene.2017.00032
+Vangimalla, S. S., Ganesan, M., Kharbanda, K. K. & Osna, N. A. Bifunctional Enzyme JMJD6 Contributes to Multiple Disease Pathogenesis: New Twist on the Old Story. Biomolecules7 (2017). https://doi.org:10.3390/biom7020041
+Zhou, J.et al.An oncogenic JMJD6-DGAT1 axis tunes the epigenetic regulation of lipid droplet formation in clear cell renal cell carcinoma. Mol Cell82, 3030–3044 e3038 (2022). https://doi.org:10.1016/j.molcel.2022.06.003
+Paschalis, A.et al.JMJD6 Is a Druggable Oxygenase That Regulates AR-V7 Expression in Prostate Cancer. Cancer Res81, 1087–1100 (2021). https://doi.org:10.1158/0008-5472.CAN-20-1807
+Yang, J.et al.Targeting Histone Demethylases in MYC-Driven Neuroblastomas with Ciclopirox. Cancer Res77, 4626–4638 (2017). https://doi.org:10.1158/0008-5472.CAN-16-0826
+Wong, M.et al.JMJD6 is a tumorigenic factor and therapeutic target in neuroblastoma. Nat Commun10, 3319 (2019). https://doi.org:10.1038/s41467-019-11132-w
+Liu, W.et al.Brd4 and JMJD6-associated anti-pause enhancers in regulation of transcriptional pause release. Cell155, 1581–1595 (2013). https://doi.org:10.1016/j.cell.2013.10.056
+Miller, T. E.et al.Transcription elongation factors represent in vivo cancer dependencies in glioblastoma. Nature547, 355–359 (2017). https://doi.org:10.1038/nature23000
+Loven, J.et al.Selective inhibition of tumor oncogenes by disruption of super-enhancers. Cell153, 320–334 (2013). https://doi.org:10.1016/j.cell.2013.03.036
+Chapuy, B.et al.Discovery and characterization of super-enhancer-associated dependencies in diffuse large B cell lymphoma. Cancer Cell24, 777–790 (2013). https://doi.org:10.1016/j.ccr.2013.11.003
+Puissant, A.et al.Targeting MYCN in neuroblastoma by BET bromodomain inhibition. Cancer Discov3, 308–323 (2013). https://doi.org:10.1158/2159-8290.CD-12-0418
+Wyce, A.et al.BET inhibition silences expression of MYCN and BCL2 and induces cytotoxicity in neuroblastoma tumor models. PLoS One8, (2013). https://doi.org:10.1371/journal.pone.0072967
+Meyers, R. M.et al.Computational correction of copy number effect improves specificity of CRISPR-Cas9 essentiality screens in cancer cells. Nat Genet49, 1779–1784 (2017). https://doi.org:10.1038/ng.3984
+Brodeur, G. M. & Bagatell, R. Mechanisms of neuroblastoma regression. Nat Rev Clin Oncol11, 704–713 (2014). https://doi.org:10.1038/nrclinonc.2014.168
+Delmore, J. E.et al.BET bromodomain inhibition as a therapeutic strategy to target c-Myc. Cell146, 904–917 (2011). https://doi.org:10.1016/j.cell.2011.08.017
+Liu, W.et al.Brd4 and JMJD6-associated anti-pause enhancers in regulation of transcriptional pause release. Cell155, 1581–1595 (2013). https://doi.org:10.1016/j.cell.2013.10.056
+Wu, G.et al.Inhibition of SF3B1 by molecules targeting the spliceosome results in massive aberrant exon skipping. RNA24, 1056–1066 (2018). https://doi.org:10.1261/rna.065383.117
+Dang, C. V. MYC, metabolism, cell growth, and tumorigenesis. Cold Spring Harb Perspect Med3 (2013). https://doi.org:10.1101/cshperspect.a014217
+Miller, D. M., Thomas, S. D., Islam, A., Muench, D. & Sedoris, K. c-Myc and cancer metabolism. Clin Cancer Res18, 5546–5553 (2012). https://doi.org:10.1158/1078-0432.CCR-12-0977
+Gordan, J. D., Thompson, C. B. & Simon, M. C. HIF and c-Myc: sibling rivals for control of cancer cell metabolism and proliferation. Cancer Cell12, 108–113 (2007). https://doi.org:10.1016/j.ccr.2007.07.006
+Wise, D. R. & Thompson, C. B. Glutamine addiction: a new therapeutic target in cancer. Trends Biochem Sci35, 427–433 (2010). https://doi.org:10.1016/j.tibs.2010.05.003
+Porter, L. D., Ibrahim, H., Taylor, L. & Curthoys, N. P. Complexity and species variation of the kidney-type glutaminase gene. Physiol Genomics9, 157–166 (2002). https://doi.org:10.1152/physiolgenomics.00017.2002
+Cassago, A.et al.Mitochondrial localization and structure-based phosphate activation mechanism of Glutaminase C with implications for cancer metabolism. Proc Natl Acad Sci U S A109, 1092–1097 (2012). https://doi.org:10.1073/pnas.1112495109
+Wang, J. B.et al.Targeting mitochondrial glutaminase activity inhibits oncogenic transformation. Cancer Cell18, 207–219 (2010). https://doi.org:10.1016/j.ccr.2010.08.009
+Gao, P.et al.c-Myc suppression of miR-23a/b enhances mitochondrial glutaminase expression and glutamine metabolism. Nature458, 762–765 (2009). https://doi.org:10.1038/nature07823
+Wise, D. R.et al.Myc regulates a transcriptional program that stimulates mitochondrial glutaminolysis and leads to glutamine addiction. Proc Natl Acad Sci U S A105, 18782–18787 (2008). https://doi.org:10.1073/pnas.0810199105
+Yuneva, M., Zamboni, N., Oefner, P., Sachidanandam, R. & Lazebnik, Y. Deficiency in glutamine but not glucose induces MYC-dependent apoptosis in human cells. J Cell Biol178, 93–105 (2007). https://doi.org:10.1083/jcb.200703099
+Masamha, C. P.et al.CFIm25 regulates glutaminase alternative terminal exon definition to modulate miR-23 function. RNA22, 830–838 (2016). https://doi.org:10.1261/rna.055939.116
+Yi, J.et al.JMJD6 and U2AF65 co-regulate alternative splicing in both JMJD6 enzymatic activity dependent and independent manner. Nucleic Acids Res45, 3503–3518 (2017). https://doi.org:10.1093/nar/gkw1144
+Li, H.et al.The landscape of cancer cell line metabolism. Nat Med25, 850–860 (2019). https://doi.org:10.1038/s41591-019-0404-8
+David, C. J., Chen, M., Assanah, M., Canoll, P. & Manley, J. L. HnRNP proteins controlled by c-Myc deregulate pyruvate kinase mRNA splicing in cancer. Nature463, 364–368 (2010). https://doi.org:10.1038/nature08697
+Rauch, J.et al.c-Myc regulates RNA splicing of the A-Raf kinase and its activation of the ERK pathway. Cancer Res71, 4664–4674 (2011). https://doi.org:10.1158/0008-5472.CAN-10-4447
+Ge, Y.et al.The splicing factor RBM25 controls MYC activity in acute myeloid leukemia. Nat Commun10, (2019). https://doi.org:10.1038/s41467-018-08076-y
+Seton-Rogers, S. RNA splicing: MYC maintains high-fidelity splicing. Nat Rev Cancer15, (2015). https://doi.org:10.1038/nrc3977
+Matera, A. G. & Wang, Z. A day in the life of the spliceosome. Nat Rev Mol Cell Biol15, 108–121 (2014). https://doi.org:10.1038/nrm3742
+Wahl, M. C., Will, C. L. & Luhrmann, R. The spliceosome: design principles of a dynamic RNP machine. Cell136, 701–718 (2009). https://doi.org:10.1016/j.cell.2009.02.009
+Pan, Q., Shai, O., Lee, L. J., Frey, B. J. & Blencowe, B. J. Deep surveying of alternative splicing complexity in the human transcriptome by high-throughput sequencing. Nat Genet40, 1413–1415 (2008). https://doi.org:10.1038/ng.259
+Wang, E. T.et al.Alternative isoform regulation in human tissue transcriptomes. Nature456, 470–476 (2008). https://doi.org:10.1038/nature07509
+Yang, X.et al.Widespread Expansion of Protein Interaction Capabilities by Alternative Splicing. Cell164, 805–817 (2016). https://doi.org:10.1016/j.cell.2016.01.029
+Liu, Y.et al.Impact of Alternative Splicing on the Human Proteome. Cell Rep20, 1229–1241 (2017). https://doi.org:10.1016/j.celrep.2017.07.025
+Daemen, A.et al.Pan-Cancer Metabolic Signature Predicts Co-Dependency on Glutaminase and De Novo Glutathione Synthesis Linked to a High-Mesenchymal Cell State. Cell Metab28, 383–399 e389 (2018). https://doi.org:10.1016/j.cmet.2018.06.003
+Redis, R. S.et al.Allele-Specific Reprogramming of Cancer Metabolism by the Long Non-coding RNA CCAT2. Mol Cell61, 520–534 (2016). https://doi.org:10.1016/j.molcel.2016.01.015
+Gong, C.et al.Sequential inverse dysregulation of the RNA helicases DDX3X and DDX3Y facilitates MYC-driven lymphomagenesis. Mol Cell (2021). https://doi.org:10.1016/j.molcel.2021.07.041
+Bosse, K. R. & Maris, J. M. Advances in the translational genomics of neuroblastoma: From improving risk stratification and revealing novel biology to identifying actionable genomic alterations. Cancer (2015). https://doi.org:10.1002/cncr.29706
+Pinto, N. R.et al.Advances in Risk Classification and Treatment Strategies for Neuroblastoma. J Clin Oncol33, 3008–3017 (2015). https://doi.org:10.1200/JCO.2014.59.4648
+Brodeur, G. M. Neuroblastoma: biological insights into a clinical enigma. Nat Rev Cancer3, 203–216 (2003). https://doi.org:10.1038/nrc1014
+Maris, J. M., Hogarty, M. D., Bagatell, R. & Cohn, S. L. Neuroblastoma. Lancet369, 2106–2120 (2007). https://doi.org:10.1016/S0140-6736(07)60983-0
+Cohn, S. L.et al.The International Neuroblastoma Risk Group (INRG) classification system: an INRG Task Force report. J Clin Oncol27, 289–297 (2009). https://doi.org:10.1200/jco.2008.16.6785
+Suh, E.et al.Late mortality and chronic health conditions in long-term survivors of early-adolescent and young adult cancers: a retrospective cohort analysis from the Childhood Cancer Survivor Study. Lancet Oncol21, 421–435 (2020). https://doi.org:10.1016/S1470-2045(19)30800-9
+Nathan, P. C.et al.Health-related quality of life in adult survivors of childhood Wilms tumor or neuroblastoma: A report from the childhood cancer survivor study. Pediatr Blood Cancer49, 704–715 (2007). https://doi.org:10.1002/pbc.20949
+Brady, S. W.et al.Pan-neuroblastoma analysis reveals age- and signature-associated driver alterations. Nat Commun11, 5183 (2020). https://doi.org:10.1038/s41467-020-18987-4
+Han, T.et al.Anticancer sulfonamides target splicing by inducing RBM39 degradation via recruitment to DCAF15. Science356 (2017). https://doi.org:10.1126/science.aal3755
+Xu, P., Duong, D. M. & Peng, J. Systematical optimization of reverse-phase chromatography for shotgun proteomics. J Proteome Res8, 3944–3950 (2009). https://doi.org:10.1021/pr900251d
+Bai, B.et al.U1 small nuclear ribonucleoprotein complex and RNA splicing alterations in Alzheimer’s disease. Proc Natl Acad Sci U S A110, 16562–16567 (2013). https://doi.org:10.1073/pnas.1310249110
+Dobin, A.et al.STAR: ultrafast universal RNA-seq aligner. Bioinformatics29, 15–21 (2013). https://doi.org:10.1093/bioinformatics/bts635
+Shannon, P.et al.Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res13, 2498–2504 (2003). https://doi.org:10.1101/gr.1239303
+
+
+
diff --git a/test/fixtures/2023.08.03.551564/2023.08.03.551564.xml b/test/fixtures/2023.08.03.551564/2023.08.03.551564.xml
new file mode 100644
index 00000000..8dfccd5c
--- /dev/null
+++ b/test/fixtures/2023.08.03.551564/2023.08.03.551564.xml
@@ -0,0 +1,511 @@
+
+
+
+
+
+BIORXIV
+
+bioRxiv
+bioRxiv
+
+
+Cold Spring Harbor Laboratory
+
+
+
+10.1101/2023.08.03.551564
+1.1
+
+
+Regular Article
+
+
+New Results
+
+
+Neuroscience
+
+
+
+Restoration of locomotor function following stimulation of the A13 region in Parkinson’s mouse models
+
+
+
+http://orcid.org/0000-0003-0628-3104
+KimLinda H
+1
+2
+*
+
+
+http://orcid.org/0000-0002-6067-4272
+LognonAdam
+1
+2
+*
+
+
+SharmaSandeep
+1
+3
+
+
+http://orcid.org/0000-0002-2856-919X
+TranMichelle A.
+3
+
+
+http://orcid.org/0000-0001-6118-813X
+ChomiakTaylor
+1
+4
+
+
+TamStephanie
+3
+
+
+McPhersonClaire
+3
+
+
+http://orcid.org/0000-0002-8361-1733
+EatonShane E. A.
+1
+3
+
+
+http://orcid.org/0000-0002-8656-2690
+KissZelma H. T.
+1
+2
+4
+
+
+http://orcid.org/0000-0002-1234-5415
+WhelanPatrick J.
+1
+3
+#
+
+Hotchkiss Brain Institute, University of Calgary, Calgary, AB, Canada, T2N4N1,
+Department of Neuroscience, University of Calgary, Calgary, AB, Canada, T2N 4N1,
+Faculty of Veterinary Medicine, University of Calgary, Calgary, AB, Canada, T2N4N1,
+Department of Clinical Neurosciences, University of Calgary, Calgary, AB, Canada, T2N 4N1
+
+
+Corresponding author Patrick J. Whelan HMRB 168, 3330 Hospital Drive NW, University of Calgary Calgary, AB T2N 4N1 Email: whelan@ucalgary.ca
+
Parkinson’s disease (PD) is characterized by extensive motor and non-motor dysfunction, including gait disturbance, which is difficult to treat effectively. This study explores the therapeutic potential of targeting the A13 region, a dopamine-containing area of the medial zona incerta (mZI) that has shown relative preservation in PD models. The A13 is identified to project to the mesencephalic locomotor region (MLR), with a subpopulation of cells displaying activity correlating to movement speed, suggesting its potential involvement in locomotor function. We show that photoactivation of this region can alleviate bradykinesia and akinetic symptoms in a mouse model of PD, revealing the presence of preserved parallel motor pathways for movement. We identified areas of preservation and plasticity within the mZI connectome using whole-brain imaging. Our findings suggest a global remodeling of afferent and efferent projections of the A13 region, highlighting the zona incerta’s role as a crucial hub for the rapid selection of motor function. Despite endogenous compensatory mechanisms proving insufficient to overcome locomotor deficits in PD, our data demonstrate that photostimulation of the A13 region effectively restores locomotor activity. The study unveils the significant pro-locomotor effects of the A13 region and suggests its promising potential as a therapeutic target for PD-related gait dysfunction.
+
+SIGNIFICANCE STATEMENT
+
This work examines the function of the A13 nucleus in locomotion, an area with direct connectivity to locomotor regions in the brainstem. Our work shows that A13 stimulation can restore locomotor function and improve bradykinesia symptoms in a PD mouse model.
Parkinson’s disease (PD) is a complex condition affecting many facets of motor and non-motor functions, including visual, olfactory, memory and executive functions (Cenci and Björklund, 2020). Due to the widespread features of PD, focusing on changes within a single pathway cannot account for all symptoms. Gait disturbance is one of the hardest to treat; pharmacological, deep brain stimulation (DBS) and physical therapies lead to only partial improvements (Nonnekes et al., 2020, 2015). While the subthalamic nucleus (STN) and globus pallidus (GPi) are common DBS targets for PD, alternative targets such as pedunculopontine nucleus (PPN) and the zona incerta (ZI) have been proposed with mixed results in improving postural and/or gait dysfunctions (Caire et al., 2013; Ferraye et al., 2010; Gut and Winn, 2015; Hamani et al., 2011; Moro et al., 2010; Nonnekes et al., 2015; Okun and Foote, 2010; Ossowska, 2019; Stefani et al., 2007; Thevathasan et al., 2018). Part of the issue with targeting the ZI with DBS strategies is the relative lack of knowledge regarding its downstream anatomical and functional connectivity with motor centres. Recent work with photoactivation of subpopulations of PPN neurons in PD models shows promise for similar ZI-focused strategies (Masini and Kiehn, 2022).
+
The ZI is recognized as an integrative hub, with roles in regulating sensory inflow, arousal, motor function, and conveying motivational states (Mitrofanis, 2005; Wang et al., 2020). As such, it is well placed to be involved in PD and has seen increased clinical and preclinical research over the last two decades (Blomstedt et al., 2018; Ossowska, 2019; Plaha et al., 2008). However, little attention has been placed on the medial zona incerta (mZI), particularly the A13, the only dopamine-containing region of the rostral ZI (Bolton et al., 2015; Kim et al., 2017; Sharma et al., 2018). Recent research in primates and mice (Peoples et al., 2012; Roostalu et al., 2019; Shaw et al., 2010) indicates that the A13 is preserved in 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-based PD models.
+
Recently, we discovered that the A13 located within the ZI projects to two areas of the mesencephalic locomotor region (MLR), the PPN and cuneiform nucleus (CnF)(Sharma et al., 2018), suggesting a role for A13 in locomotor function. Indeed, in vivo photometry recordings from calcium/calmodulin-dependent protein kinase IIα (CaMKIIα) populations in the rostral ZI, which includes the A13 nucleus, show a subpopulation of cells whose activity correlates with movement speed (Li et al., 2021). Since this region projects to the MLR, it is a potential parallel motor pathway to target for gait improvement. Photoactivation of glutamatergic MLR neurons alleviates motor deficits in the 6-OHDA mouse model (Fougère et al., 2021; Masini and Kiehn, 2022). Phenomena such as kinesia paradoxa (Glickstein and Stein, 1991) in PD patients support the existence of preserved parallel motor pathways that can be engaged in particular circumstances to produce normal movement.
+
Further evidence supporting the importance of parallel motor pathways in PD includes those reporting functional alterations in A13 (Hoffman et al., 1997; Périer et al., 2000). Nigrostriatal lesions affect A13 cellular function and lead to anatomical remodeling in monoaminergic brain regions (Braak et al., 2003; Kish et al., 2008; Lim et al., 2009; Perez-Lloret and Barrantes, 2016; Roostalu et al., 2019; Scatton et al., 1983; Zweig et al., 1989). The A13 connectome encompasses the cerebral cortex (Mitrofanis and Mikuletic, 1999), central nucleus of the amygdala (Eaton et al., 1994), thalamic paraventricular nucleus (Li et al., 2014), thalamic reuniens (Sita et al., 2007; Venkataraman et al., 2021), MLR(Sharma et al., 2018), superior colliculus (SC) (Bolton et al., 2015), and dorsolateral periaqueductal grey (PAG) (Messanvi et al., 2013; Sita et al., 2007), making the A13 an important hub for goal-directed locomotion (Choi and McNally, 2017; Eaton et al., 1994; Messanvi et al., 2013; Mok and Mogenson, 1986; Moriya et al., 2020; Ogundele et al., 2017; Manjit K. Sanghera et al., 1991; M. K. Sanghera et al., 1991; Sita et al., 2007; Venkataraman et al., 2021).
+
Based on the role of the A13 in gait and, specifically, as a possible target to improve gait in PD, we investigated the therapeutic potential of photoactivating a tightly circumscribed region targeting a small region containing mainly the A13 and a small area of the mZI, which we term A13 region throughout the manuscript. We identified areas of preservation and plasticity within the mZI connectome using whole-brain imaging techniques. Photoactivation of the A13 region rescued bradykinetic and akinetic symptoms in a mouse model of 6-hydroxydopamine (6-OHDA) mediated unilateral nigrostriatal degeneration. Because the zona incerta is a hub for the rapid selection of motor function, we then mapped the input and output patterns of the region. We found evidence of a global remodeling of afferent and efferent projections of the A13 region. While endogenous compensatory mechanisms from the remaining but remodelled A13 region connectome were inadequate in overcoming locomotor deficits observed in 6-OHDA mice, photostimulation of the A13 region restored locomotor activity. These data demonstrate that the A13 region produces powerful pro-locomotor effects in normal and PD mouse models. Moreover, PD-related bradykinesia is ameliorated with A13 region photoactivation in the presence of remodelling of the A13 region connectome. Some of these data have been published in abstract form (L. Kim et al., 2021).
+
+
+RESULTS
+
+Unilateral 6-OHDA mouse model has robust motor deficits
+
The overall experimental design is illustrated in Figure 1A, along with a schematic in Figure 1B showing injections of 6-OHDA in the medial forebrain bundle and AAVDJ-CaMKIIα-ChR2 virus into the medial zona incerta (mZI). We confirmed substantia nigra pars compacta (SNc) degeneration in a well-validated unilateral 6-OHDA-mediated Parkinsonian mouse model (Thiele et al., 2012). The percentage of tyrosine hydroxylase (TH+) cell loss normalized to the intra-animal contralesional side was quantified. 6-OHDA produced a significant lesion that decreased TH+ neuronal SNc populations. As previously reported (Boix et al., 2015), the SNc ipilesional to the 6-OHDA injection (n = 10) showed major ablation of the TH+ neurons compared to sham animals (Figure 1C and D: n = 11).
+
+
+
Experimental design and confirmation of unilateral TH<sup>+</sup> depletion in the SNc via 6-OHDA lesion.
+
(A) Illustration of experimental timeline. (B) Dual ipsilateral stereotaxic injection into the MFB and A13 region. (C) TH+ cells in SNc of sham (top) compared to 6-OHDA injected mouse (bottom). Magnified areas outlined by yellow squares are shown on the right. (D) Unilateral injection of 6-OHDA (6-OHDA ChR2: n = 5, 6-OHDA eYFP: n = 5) into the MFB resulted in greater percentage of TH+ loss compared to sham in the SNc (sham ChR2: n = 7, sham eYFP: n = 5, three-way MM ANOVAs), regardless of virus type (F1,18 = 104.4, p < .001). ***p < .001. Error bars indicate SEMs.
+
+
+
+
+A13 region photoactivation generates pro-locomotor behaviors in the open field
+
6-OHDA lesions are characterized as generating bradykinetic and akinetic phenotypes in the open field (Li et al., 2022; Magno et al., 2019; Masini and Kiehn, 2022; Sanders and Jaeger, 2016). To understand the impact of A13 region photoactivation on locomotion in sham and PD model mice, on-target localization of ferrule above the A13 region, centered on the mZI, along with YFP reporter expression, was confirmed (Figure 2) in mice given sham or 6-OHDA injections. Corroborating the post hoc targeting, we found evidence for c-Fos in neurons within the A13 region in photostimulated ChR2 mice (Figure 2). Before post hoc analysis, mice were monitored in the open field test (OFT), where the effects of the 6-OHDA lesion were apparent, with 6-OHDA lesioned animals demonstrating far less movement, fewer bouts of locomotion, and less time engaging in locomotion in the OFT (Figure 2A-E). Notably, photoactivation of the A13 region often generated dramatic effects, with mice showing a distinct increase in locomotor behavior (Figure 3A, Movie S1 & Movie S2). Both sham and 6-OHDA ChR2 mice showed a significant increase in locomotor distance travelled during periods of photoactivation (Figure 3B, p = 0.005). One sham animal showed grooming behavior on stimulation and was excluded from the analysis.
+
+
+
<italic>Post hoc</italic> c-Fos expression and targeting of the mZI and A13.
+
(A) Diagram showing the A13 DAergic nucleus in dark magenta encapsulated by the ZI in light magenta. The fibre optic tip is outlined in red. Atlas image adapted from the Allen Brain Atlas (Goldowitz, 2010). (B) Tissue images were obtained from 6-OHDA ChR2 animals around bregma −1.22 mm, and (C) a 6-OHDA eYFP animal more caudally around bregma −1.46 mm. Images show the distribution of DAPI (blue), eYFP (green), c-Fos (yellow), and TH (magenta). Landmarks are outlined in white (3V: third ventricle; mtt: mammillothalamic tract), and the optic cannula tip is shown in red. Higher magnification images of the A13 DAergic nucleus are outlined by the yellow boxes in a 6-OHDA ChR2 animal (D) and a 6-OHDA eYFP animal (E). Scale bars are set to 350 μm. Images show isolated channels in the top rows of the respective groups: eYFP (i), TH (ii), and c-Fos (iii). Merged channels for eYFP and c-Fos (iv), TH and c-Fos (v), and a merge of all four channels (vi) are presented in the bottom rows of their respective groups. White arrowheads in the merged images highlight overlap in merged markers. Red arrows show triple colocalization of eYFP, c-Fos and TH. (Dvi) contains a magnified example of triple-labelled neurons, as highlighted in the yellow box. Scale bars are set to 50 μm.
+
+
+
+
+
Ispilesional photoactivation of the A13 region in a unilateral 6-OHDA mouse model rescues motor deficits.
+
(A) Schematic of open field experiment design and example traces for open field testing (1 min) with unilateral photoactivation of the A13 region. (B-E) Effects of photoactivation on open field metrics for sham eYFP (n = 5), sham ChR2 (n = 6), 6-OHDA eYFP (n = 5), and 6-OHDA ChR2 (n = 5) groups (three-way MM ANOVAs, post hoc Bonferroni pairwise). Photoactivation increased in the ChR2 groups: (B) distance travelled (ChR2 vs. eYFP: p = 0.005), (C) locomotor bouts (ChR2 vs. eYFP: p = 0.005), (D) duration of locomotion in the open field (ChR2 vs. eYFP: p = 0.005), and (E) animal movement speed (ChR2 vs. eYFP: p < 0.001). (F-I) Group averaged instantaneous velocity graphs showing no increase in a sham eYFP (F) or 6-OHDA eYFP mouse (H), with increases in velocity during stimulation in a sham ChR2 (G) and 6-OHDA ChR2 (I) mouse. (J) The graph presents animal rotational bias using the turn angle sum. There was a significant increase in 6-OHDA ChR2 rotational bias during A13 region photoactivation (6-OHDA ChR2 vs. 6-OHDA eYFP: p < 0.001). (K) Diagram depicting the pole test. A mouse is placed on a vertical pole facing upwards. The time for release is taken as the experimenter removes their hand from the animal’s tail. (L, M) Graphs showing the response of animals to photoactivation of the A13 region while performing the pole test. (L) A13 region photoactivation also led to shorter total descent time in ChR2 compared to eYFP mice (ChR2 vs. eYFP: p = 0.004), and (M) 6-OHDA ChR2 mice showed a greater reduction in descent time compared to sham ChR2 (6-OHDA ChR2 vs. sham ChR2: p = 0.012; 6-OHDA ChR2: n = 5; sham ChR2: n = 7). ***p < .001, **p < .01, *p < .05. Bonferroni’s post hoc comparisons between 6-OHDA ChR2 and sham eYFP, sham ChR2, and 6-OHDA eYFP at stim time point as a, b, and c respectively. Error bars indicate SEMs.
+
+
+
We tested whether photoactivation led to a single bout of locomotion or if there was an overall increase in bouts, signifying that animals could repeatedly initiate locomotion following photoactivation. Mice in the ChR2 groups demonstrated an increase in the number of locomotion bouts with photoactivation, indicating a greater ability to start locomotion from rest, and that photoactivation was not eliciting a single prolonged bout (Figure 3C, p = 0.005). When we examined each bout of locomotion, photoactivation increased the total duration of locomotion (Figure 3D, p = 0.005). There was a refractory decrease in the distance travelled by the sham ChR2 animal group (Figure S1A), which was not evident for the 6-OHDA cohort (Figure S1B). To control for this, we compared the pre-timepoints to the baseline one-minute averages to ensure that the animal locomotion distance travelled returned to a stable state before stimulation was reapplied (Figure S1C, p = 0.783).
+
Next, we examined the reliability of photoactivation to initiate locomotion. The percentage of trials with at least one bout of locomotion was compared for the pre-and stim time points. 6-OHDA ChR2 animals showed a reliable pro-locomotion phenotype with A13 region photoactivation (Figure S2A: p = 0.042). As was expected in the control 6-OHDA eYFP group, there was no effect of photoactivation on the probability of engaging in locomotion (p = 0.713).
+
Animal movement speed also factors into the total distance travelled measure and can be discussed in regard to a bradykinetic phenotype in 6-OHDA lesioned mice (Magno et al., 2019; Masini and Kiehn, 2022; Sanders and Jaeger, 2016). Using instantaneous animal movement speeds that exceeded 2 cm/s as per Masini & Kiehn (2022), we plotted instantaneous speed (Figure 3F-I) and analyzed one-minute bins (Figure 3E). As was expected, 6-OHDA lesioned animals had lower movement speeds than sham control animals (p < 0.001). One animal from the 6-OHDA eYFP group was excluded because it did not meet the speed threshold during recording. Both the 6-OHDA ChR2 and sham ChR2 groups displayed increases in average speed during photostimulation (Figure 3E, p < 0.001). When we examined the time taken to initiate locomotion, there was no significant difference between sham or 6-OHDA ChR2 groups (Figure S2B).
+
+
+Photoactivation of the A13 region increases ipsilesional turning in the open field test
+
Unilateral 6-OHDA lesions drive asymmetric rotational bias (Boix et al., 2015; Li et al., 2022; Magno et al., 2019; Thiele et al., 2012). We were interested in whether this persisted with stimulation and noted upon observation of photoactivation that animals appeared to have increased ipsilesional rotation. As observed, 6-OHDA ChR2 animals had an increase in turn angle sum (TAS), indicating an increase in their rotational bias with photoactivation in the ipsilesional direction (Figure 3J, p < 0.001). As expected, 6-OHDA eYFP animals showed consistent rotational bias throughout time. The rotational bias of sham ChR2 was also compared to determine whether the increased rotational bias was due to photoactivation or the interaction of photoactivation and the lesion. The sham ChR2 group showed no significant change in TAS with photoactivation (Figure 3J, p > 0.05). Next, we examined whether the increased turning angle sum in the 6-OHDA ChR2 group was observed during periods of locomotion. When the TAS was calculated only during periods of locomotion, the rotational bias in the animal orientation in the 6-OHDA ChR2 animals was not observed (p = 0.286).
+
+
+Skilled vertical locomotion is improved in the pole test with photoactivation of the A13 region
+
The pole test is a classic 6-OHDA behavioral paradigm (Figure 3K) that involves skilled locomotor abilities for an animal to turn and descend a vertical pole (Matsuura et al., 1997; Ogawa et al., 1985). Improvements in function can be inferred if the time taken to complete the test decreases (Matsuura et al., 1997; Ogawa et al., 1985). 6-OHDA mice demonstrated significantly greater descent times than sham mice (p < 0.001). Photoactivation of the A13 region reduced descent times for both 6-OHDA and sham groups on the pole test (Figure 3L, p = 0.004, Movie S3). Neither of the eYFP groups showed any changes in the time to complete the pole test.
+
To further understand the effects of photoactivation on the ability of mice to descend the pole, the time taken for mice to descend after turning was analyzed to remove any influence of animals spending time investigating their environment on the top of the pole. While all groups showed reduced total pole test descent time with photoactivation, considering just the time to descend from turn alone, there was a larger improvement with A13 region photoactivation in the 6-OHDA ChR2 mice compared to sham ChR2 mice (Figure 3M: p = 0.012). These results indicate that photoactivation has the effect of reducing bradykinesia by improving the ability of mice to descend the pole during the PT.
+
+
+Dopaminergic Cells in the A13 region are preserved in the unilateral 6-OHDA mouse model
+
While photoactivation of the A13 region promoted locomotor activity in both sham and 6-OHDA mice, there were differences in speed and directional bias. We hypothesized that this may be due to changes in the A13 region connectome since there is evidence of changes in firing and metabolic activity in the region (Périer et al., 2000). Therefore, we utilized whole brain imaging approaches (Hansen et al., 2020; Zhan et al., 2021) to examine changes in the connectome following 6-OHDA lesions of the nigrostriatal region.
+
Using whole brain imaging, as expected, TH+ cells in SNc were more vulnerable to the 6-OHDA neurotoxin than the ventral tegmental area (VTA) and A13 (Figure S3A-F). 6-OHDA-treated mice showed a significantly greater percentage of TH+ cell loss in SNc compared to the VTA and A13 (VTA vs. SNc: p = 0.003; A13 vs. SNc: p = 0.005). In contrast, sham animals showed no significant difference in TH+ cell loss across SNc, VTA and A13 (Figure S3G, p > 0.05). Thus, similar to that observed in the human brain of Parkinsonian patients (Matzuk and Saper, 1985), there is a remarkable preservation of dopaminergic cells in the A13 after nigrostriatal degeneration in the 6-OHDA mouse model of PD.
+
+
+Large-scale changes in the A13 region connectome following 6-OHDA-mediated unilateral nigrostriatal degeneration
+
Although photoactivation had benefits in restoring speed in 6-OHDA mice (Figure 3), circling behavior was increased, suggesting additional changes that may reflect connectome alterations. We examined the changes in the input and output of the A13 by co-injecting anterograde (AAV8-CamKII-mCherry) and retrograde AAV (AAVrg-CAG-GFP) tracers into the A13 nucleus (Keith B. J. Franklin and Paxinos, 2008). The injection core and spread were determined in the rostrocaudal direction from the injection site (Figure S4). To examine whether unilateral nigrostriatal degeneration resulted in changes in the organization of inputs and outputs from the A13, we first visualized interregional correlations of afferent and efferent proportions for each condition using correlation matrices (Fig 4A and B; 251 regions in a pairwise manner). Correlation matrices were organized using the hierarchical anatomical groups from the Allen Brain Atlas (Figure 4C). To minimize the influence of experimental variation on the total labeling of neurons and fibers, the afferent cell counts or efferent fiber areas in each brain region were divided by the total number found in a brain to obtain the proportion of total inputs and outputs. The data were normalized to a log10 value to reduce variability and bring brain regions with high and low proportions of cells and fibers to a similar scale (Kimbrough et al., 2020). Comparing the afferent and efferent proportions in a pairwise manner between mice showed good consistency with an average correlation of 0.91 ± 0.02 (Spearman’s correlation, Figure S5).
+
+
+
Unilateral nigrostriatal degeneration leads to large-scale changes in the organization of the A13 region afferent and efferent distributions across the neuraxis.
+
We used correlation matrices to summarize any observable patterns in the distribution patterns of inputs and outputs of the A13 region. A correlation matrix was calculated by correlating the proportion of input from one brain region to another in a pairwise manner across 251 brain regions delimited by registration with Allen Brain Atlas. If two brain regions among mice (eg. brain regions A and B) contribute a similar input, they are highly correlated (A). Using the color legend showing various correlation strengths, the intersecting box in the matrix in this example will be colored dark red (B). If no relationship is found between contributions from two brain regions, the intersecting box will be colored yellow. If the contribution from one brain region was negatively correlated with another brain region among mice, then the intersecting box will be colored blue. The afferent distribution pattern in the sham displayed a higher level of inter-regional correlation between brain regions (C) than 6-OHDA injected mice (D). Indeed, two distinct bands of anti-correlated afferent regions were identified in the 6-OHDA injected mice (see black boxes in D). These two bands arose from the cortical plate subregions (motor, sensory, visual, and prefrontal), and striatal and pallidal subregions showing distinct inputs compared to the rest of the neuraxis. In contrast, the projection patterns of A13 efferents displayed a higher level of inter-regional correlation between brain regions following a unilateral nigrostriatal degeneration (F) compared to sham (E). In sham, proportions of A13-cortical/ striatal efferents were negatively correlated to A13-pallidal/ thalamic/ hypothalamic/ midbrain efferents (see black boxes in E). However, these distinct projection patterns disappeared following nigrostriatal degeneration, suggesting A13 efferent distributions becoming more distributed across the neuraxis.
+
+
+
We observed changes in projection patterns between the sham and the 6-OHDA group. A correlation matrix was used to quantify the relationship of input from brain regions in a pairwise manner through the neuraxis. Overall, afferents onto the A13 in sham animals displayed a higher interregional correlation between brain regions than 6-OHDA-injected mice. Specifically, correlation coefficients of 0.10, 0.30, and 0.50 or larger represent weak, moderate, and strong correlations, respectively (Cohen, 1988). Compared to sham (Figure 4C), in 6-OHDA injected mice (Figure 4D), afferent contributions from two clusters of brain subregions became more dissimilar (anti-correlated) to the rest of the neuraxis: 1) cortical plate subregions (motor, sensory, visual, and prefrontal), and 2) striatal, and pallidal subregions compared to sham (boxed blue areas in Figure 4D). These data suggest that afferents from several regions showed a coordinated reduction in afferent density onto the A13, except contributions from cortical plate (motor, sensory, visual, and prefrontal), striatal, and pallidal subregions that were positively correlated. In other words, contributions from the cortical plate (motor, sensory, visual, and prefrontal), striatal and pallidal subregions are positively correlated, but compared to the rest of the brain, they are anti-correlated. This suggests a greater afferent input onto A13 from the cortical plate (motor, sensory, visual, and prefrontal), striatal and pallidal subregions than other regions after 6-OHDA lesions.
+
In marked contrast, the projection patterns of A13 efferents exhibited a higher level of interregional correlation between brain regions following a unilateral nigrostriatal degeneration compared to sham. In the sham condition, the A13 connectome is biased towards cortical and striatal regions compared to pallidal, thalamic, hypothalamic, midbrain efferents. This is shown by a broad negative correlation between these two large groups (Figure 4E). However, these broad anti-correlations disappear following nigrostriatal degeneration (Figure 4F). These data indicate that the A13 efferent connectome is less refined following nigrostriatal degeneration.
+
+
+Differential remodeling of A13 region connectome ipsi- and contra-lesion following 6-OHDA-mediated nigrostriatal degeneration
+
The distributions of the A13 connectome in sham animals served as the basis for an in-depth comparison of the preservation and plasticity of A13 afferents and efferents in 6-OHDA mouse models (Figure 5). We observed a global remodeling of A13 afferents and efferents following unilateral nigrostriatal degeneration (Figure 5A, B; E, F) that was differentially expressed across the neuraxis (Figure 4D, H). The ipsilesional side showed more downregulated areas (Figure 5D: see also example traces). These downregulated areas were focused within the cortical plate and cortical subplate regions. 6-OHDA injections also downregulated A13 afferent densities from the striatum, pallidum, thalamus and medulla. While 6-OHDA mainly downregulated ipsilesional A13 afferent densities, the hypothalamus (including ZI), midbrain, pons, and cerebellum had increased A13 afferent densities.
+
+
+
Differential remodeling of A13 region connectome following a unilateral nigrostriatal degeneration.
+
The distributions of the A13 connectome in sham served as a basis for an in-depth comparison against 6-OHDA mouse models. Example registered slices (using WholeBrain software 64 with light-sheet data, 2X objective, 4X optical zoom) at rostral areas show changes in sham (A) afferents compared to 6-OHDA lesioned animals (B). Graph showing major brain regions contributing afferents to A13 in sham mice (C). The graph illustrates the change in the proportion of afferents in 6-OHDA compared to sham mice (D). Representative registered slices showing sham proportions of efferents in sham (E) compared to 6-OHDA mice (F). The magnified black box section displays an example of mCherry+ fibers (left) segmented using Ilastik and ImageJ software (right). Graph showing major brain regions receiving efferents from the A13 in sham mice (G). The graph illustrates the change in the proportion of efferents in 6-OHDA compared to sham mice (H). Error bars represent SEMs. Anterograde and retrograde viruses were injected into the ipsilesional A13 (see methods). Abbreviations from Allen Brain. Atlas: CTXpl (cortical plate), CTXsp (cortical subplate), STR (striatum), PAL (palladium), TH (thalamus), HYP (hypothalamus), P (pons), MB (midbrain), MY (Medulla), and CB (cerebellum).
+
+
+
The intact, contralesional side showed more upregulated regions across the neuraxis, suggesting compensatory upregulation in the unilateral 6-OHDA model (Figure 5D). The cortical plate, striatum, and cortical subplate were the top three upregulated contralesional regions. In contrast, contralesional A13 afferent densities from the pallidum and thalamus were spared and upregulated. Thus, compensatory upregulation of A13 afferent density from these regions appeared lateralized from the intact, contralesional side. Furthermore, bilateral compensatory upregulation of A13 afferents was observed from the hypothalamus (including ZI), midbrain and pons.
+
The A13 efferents were more downregulated on the ipsilesional side (Figure 5H). However, the downregulation was focused within the isocortical, striatal and cortical subplate regions. Remodeling on the contralesional efferent projection patterns closely followed the changes seen with the afferents, except for projections onto the thalamic and midbrain regions. The A13 efferents onto thalamic regions were bilaterally upregulated. Also, the A13-midbrain efferents were upregulated ipsilesionally.
+
+
+
+DISCUSSION
+
Our work demonstrates robust pro-locomotor effects induced by photoactivation of the A13 region in lesioned and sham mice. Photoactivation during the OFT increased locomotion distance travelled, the duration of locomotion, and speed in both sham and 6-OHDA mice. Uniquely, the 6-OHDA group had increases in the number of locomotor bouts which resembled the normal number of bouts observed in healthy mice at baseline. Bradykinesia in 6-OHDA mice was substantially improved following photoactivation. We found extensive input and output connectivity of the A13, which was remodeled following nigrostriatal lesions. Afferent input patterns displayed a marked reduction in interregional correlation across brain regions in 6-OHDA mice, while efferent projections increased. This demonstrates the impact of nigrostriatal lesions on dopaminergic-containing regions outside the nigrostriatal zone. These findings highlight the pro-locomotory effect, therapeutic potential, and plasticity of the A13.
+
+The role of the A13 region in locomotion in sham mice
+
We provide the first direct evidence of the photoactivation of the A13 being sufficient in driving locomotion. It is now evident that the pro-locomotor function of ZI extends further rostrally than previous work in caudal ZI (cZI) indicates (Mitrofanis, 2005): photoactivation of cZI neurons increased animal movement speed in prey capture (Zhao et al., 2019) and active avoidance (Hormigo et al., 2020). Previous data targeting the mZI region, including somatostatin (SOM+), calretinin (CR+), and vGlut2+ neurons, did not change locomotor distance travelled in the OFT (Li et al., 2021). In our work, there may be a combinatorial effect of multiple populations being photostimulated or targeting more medial populations in the ZI. Our results are consistent with in vivo calcium dynamics from CaMKIIα+ rostral ZI cells, which overlap the A13 showing subpopulations whose activity correlates with either movement speed or anxiety-related locations (Li et al., 2021). Enhancement of the A13 activity appears to modulate locomotor activity in naive mice differently from more lateral GABAergic ZI populations (dorsal and ventral ZI). Microinjection of GABAA receptor agonists, muscimol (Wardas et al., 1988) or etomidate (Chen et al., 2023), into the ZI either evokes severe catalepsy or a significant reduction in locomotor distance and velocity, respectively. Suppression of GABAergic ZI activity can either increase locomotion by microinjection of GABAA receptor antagonist bicuculline (Périer et al., 2002) or induce bradykinesia and akinesia by chemogenetic or optogenetic inhibitions in healthy naive mice (Chen et al., 2023).
+
Furthermore, our previous work showing A13 projections to the MLR is consistent with our observed photoactivation effects. We, and others, demonstrated that the A13 contains DA neurons, which may contribute to the observed effects, possibly via D1/5 receptor activation (Ryczko et al., 2016). A13 region photoactivation produces increased locomotor speed averaging 13 cm/s and improves descent times on the pole test. The enhanced ability to perform the pole test, which requires the animal to grasp the vertical pole to descend safely without falling, provides further evidence for the role of A13 region neurons in movements. Since A13 stimulation did not alter coordination during the task, it suggests a complex behavioral role consistent with its upstream location from the brainstem and extensive connectome. A13 photoactivation increases animal speed, duration, and distance travelled. Interestingly, the latency in A13 to observe increases in ongoing animal speed or to initiate locomotion is long in both 6-OHDA and sham mice (∼ 15 and 5 seconds, respectively). Second-long delays are often typical of sites upstream of the cuneiform, such as the dlPAG, which has a delay of several seconds (Tsang et al., 2021). The delays in locomotor initiation and context-specific integration following stimulation of upstream CnF targets may offer a therapeutic advantage for overcoming gait dysfunction.
+
+
+Photoactivation of the A13 reduces bradykinesia and akinesia in mouse PD models
+
Several studies have focused on the basal ganglia by targeting the subthalamic nucleus (Gradinaru et al., 2009; Yoon et al., 2014), endopeduncular nucleus (Moon et al., 2018; Yoon et al., 2020), SNc (Kravitz et al., 2010), striatum (Bordia et al., 2016; Ryan et al., 2018), cZI (Li et al., 2022), and motor cortex (Magno et al., 2019; Sanders and Jaeger, 2016; Valverde et al., 2020), or their projections in mouse models of PD. MLR subpopulations have been explored as a target for PD DBS with mixed results (Fougère et al., 2021; Masini and Kiehn, 2022). Recently, Li et al. found that cZI glutamatergic neurons were overactive after administering 6-OHDA into the striatum, and photoinhibition rescued the motor deficits (Li et al., 2022). Motor deficits in a 6-OHDA-induced PD mouse model were also ameliorated by chemogenetic and optogenetic activation of dorsal and ventral ZI GABAergic neurons (Chen et al., 2023). Here, we introduce a novel subpopulation in ZI (A13 cells) whose photoactivation rescued bradykinetic and akinetic deficits observed in the 6-OHDA mice.
+
Our work shows that A13 projections are affected at cortical and striatal levels following 6-OHDA, consistent with our observed changes in locomotor function. Over 28 days, there was a remarkable change in the afferent and efferent A13 regional connectome, despite the preservation of TH+ ZI cells. This is consistent with previous reports of widespread connectivity of the ZI (Mitrofanis, 2005). The preservation of A13 is expected since A13 lacks DAT expression (Bolton et al., 2015; Negishi et al., 2020; Sharma et al., 2018) and is spared from DAT-mediated toxicity of 6-OHDA (Dauer and Przedborski, 2003; Konnova et al., 2018; Simola et al., 2007). While A13 cells are spared following nigrostriatal degeneration, our work demonstrates its connectome is rewired. The ipsilateral afferent projections were markedly downregulated, while contralesional projecting afferents showed upregulation. In contrast, efferent projections showed less downregulation in the cortical subplate regions and bilateral upregulation of thalamic and hypothalamic efferents. Similar timeframes for anatomical and functional plasticity affecting neurons and astrocytes following an SNc or MFB 6-OHDA have been previously reported (Bosson et al., 2015; Perović et al., 2005; Requejo et al., 2020). Human PD brains that show degeneration of the SNc have a preserved A13 region, suggesting that our model, from this perspective, is externally valid (Matzuk and Saper, 1985).
+
Combined with photoactivation of the A13 region, we provide evidence for plasticity following damage to SNc. A previous brain-wide quantification of TH levels in the MPTP mouse model identified additional complexity in regulating central TH expression compared to conventional histological studies (Roostalu et al., 2019). Roostalu et al. reported decreased SNc TH+ cell numbers without a significant change in TH+ intensity in SNc and increased TH+ intensity in limbic regions such as the amygdala and hypothalamus (Roostalu et al., 2019). Likewise, we found no significant change in A13 TH+ cell counts. Still, there was a downstream shift in the distribution pattern of A13 efferents following nigrostriatal degeneration with a pullback on outputs to cortical and striatal subregions. This suggests A13 efferents are more distributed across the neuraxis than in sham mice. The preserved A13 efferents could provide compensatory dopaminergic innervation with collateralization mediated contralesionally and, in some subregions, ipsilesionally to increase the availability of extracellular dopamine. Considering A13-MLR efferents (Sharma et al., 2018) that remain preserved, photoactivation of glutamatergic MLR neurons alleviates motor deficits in the 6-OHDA mouse model (Fougère et al., 2021; Masini and Kiehn, 2022), and photoactivation of A13 somata promotes locomotion in 6-OHDA mice - hypotheses that warrant further investigation.
+
Several A13 efferent targets could be responsible for rotational asymmetry. In a unilateral 6-OHDA model, ipsiversive circling behaviour is indicative of intact striatal function on the contralesional side (Carey, 1991; Schwarting et al., 1991; Ungerstedt, 1971; Zetterström et al., 1986). Instead, the predictive value of a treatment is determined by contraversive circling mediated by increased dopamine receptor sensitivity on the ipsilesional striatal terminals (Costall et al., 1976; Lane et al., 2006). Thus, our data suggest that photoactivation of ipsilesional A13 has an overall additive effect on ipsiversive circling and represents a gain of function on the intact side that contributes to the magnitude of overall motor asymmetry against the lesioned side. Since A13 cells are preserved in PD, future therapies could use bilateral stimulations optimized for each side to minimize the overall motor asymmetry while ameliorating bradykinesia and akinesia.
+
With the induction of a 6-OHDA lesion, there is a change in the A13 connectome, characterized by a reduction in bidirectional connectivity with ipsilesional cortical regions. In rodent models, the motor cortices, including the M1 and M2 regions, can shape rotational asymmetry (Gradinaru et al., 2009; Magno et al., 2019; Sanders and Jaeger, 2016; Valverde et al., 2020). Activation of M1 glutamatergic neurons increases the rotational bias (Valverde et al., 2020), while M2 neuronal stimulation promotes contraversive rotations (Magno et al., 2019). Our data suggest that A13 photoactivation may have resulted in the inhibition of glutamatergic neurons in the contralesional M1. An alternative possibility is the activation of the contralesional M2 glutamatergic neurons, which would be expected to induce increased ipsilesional rotations (Magno et al., 2019). The ZI could generate rotational bias by A13 modulation of cZI glutamatergic neurons via incerto-incertal fibres (Ossowska, 2019; Power and Mitrofanis, 1999), which promotes asymmetries by activating the SNr (Li et al., 2022). The incerto-incertal interconnectivity has not been well studied, but the ZI has a large degree of interconnectivity (Sharma et al., 2018; Tsang et al., 2021) along all axes and between hemispheres (Power and Mitrofanis, 1999). However, this may only contribute minimally given that unilateral photoactivation of the A13 cells in sham mice failed to produce ipsiversive turning behavior while unilateral photoactivation of cZI glutamatergic neurons in sham animals was sufficient in generating ipsiversive turning behavior (Li et al., 2022). Another possibility involves the A13 region projections to the MLR. With the unknown downstream effects of A13 photoactivation, there may be modulation of the PPN neurons responsible for this turning behavior (Masini and Kiehn, 2022). The thigmotaxic behaviors suggest some effects may be mediated through dlPAG and CnF (Tsang et al., 2021), and recent work suggests the CnF as a possible therapeutic target (Fougère et al., 2021; Noga and Whelan, 2022). Since PD is a heterogeneous disease, our data provide another therapeutic target providing context-dependent relief from symptoms. This is important since PD severity, symptoms, and progression are patient-specific.
+
+
+Towards a preclinical model
+
To facilitate future translational work applying DBS to this region, we targeted the A13 region using AAV8-CamKII-mCherry viruses. The CaMKIIα promoter virus is beneficial because it is biased towards excitatory cells (Haery et al., 2019), narrowing the diversity of transfected A13 region neurons and in our hands, the viral spread was contained within the A13 region. Optogenetic strategies have been used to activate retinal cells in humans, partially restoring visual function and providing optimism that AAV-based viral strategies can be adapted in other human brain regions (Sahel et al., 2021). A more likely possibility for stimulation of deep nuclei is that DREADD technology could be adapted, which would not require any implants, but this remains a longer-term possibility. In the short-term, our work suggests that the A13 is a possible target for DBS. Gait dysfunction in PD is particularly difficult to treat, and indeed when DBS of subthalamic nucleus is deployed, a mixture of unilateral and bilateral approaches have been used (Lizarraga et al., 2016), along with stimulation of multiple targets (Stefani et al., 2007). This represents the heterogeneity of PD and underlines the need for considering multiple targets. In this regard, the identification of non-canonical dopaminergic pathways for the direct control of locomotion is promising (Figure 6). Our work highlights the A13 as a possible target, likely used in context and in concert with the activation of other identified targets.
+
+
+
Comparing descending dopamine pathways for locomotor control.
+
Simplified connectivity map for the 3 dopamine pathways. The first pathway is the classical VTA/SNc projection to the striatum, and the SNr/GPi projects to the MLR. The VTA/SNc also directly projects to the MLR (Ryczko et al., 2016). The mZI/A13 region projects dopaminergic projections to the MLR (Sharma et al., 2018). Canonical pathways are in black, while non-canonical pathways are in red.
+
+
+
+
+Limitations
+
Currently, few PD animal models are available that adequately model the progression and the extent of SNc cellular degeneration while meeting the face validity of motor deficits (Dauer and Przedborski, 2003; Konnova et al., 2018). While the 6-OHDA models fail to capture the age-dependent chronic degeneration observed in PD, it provides additional advantages in providing robust motor deficits with acute degeneration and identifying compensatory changes compared to the unlesioned side. Moreover, it resembles the unilateral onset (Hughes et al., 1992) and persistent asymmetry (Lee et al., 1995) of motor dysfunction in PD. Another option could be the MPTP mouse model, which offers the ease of systemic administration and translational value to primate models; however, the motor deficits are variable and lack the asymmetry observed in human patients (Hughes et al., 1992; Jagmag et al., 2015; Lee et al., 1995; Meredith and Rademacher, 2011). Despite these limitations, the neurotoxin-based mouse models, such as MPTP and 6-OHDA, offer greater SNc cell loss than genetic-based models; in the case of the 6-OHDA model, it captures many aspects of motor dysfunctions in PD (Dauer and Przedborski, 2003; Jagmag et al., 2015; Konnova et al., 2018; Simola et al., 2007).
+
+
+Conclusions
+
Parkinson’s disease involves areas outside the classic nigrostriatal axis. Our work demonstrates that the A13 region drives locomotor activity and rescues bradykinetic and akinetic deficits caused by dysfunctional DAergic transmission in the basal ganglia. We show that A13 region-evoked locomotion has therapeutic potential for improving gait in advanced PD. Widespread remodelling of the A13 region connectome is critical to our understanding of the effects of dopamine loss in PD models. In summary, our findings support an exciting role for the A13 region in locomotion with demonstrated benefits in a mouse PD model and contribute to our understanding of heterogeneity in PD.
+
+
+
+MATERIALS AND METHODS
+
+Animals
+
All care and experimental procedures were approved by the University of Calgary Health Sciences Animal Care Committee (Protocol #AC19-0035). C57BL/6 male mice 49 - 56 days old (weight: M = 31.7 g, SEM = 2.0 g) were group-housed (≤ four per cage) on a 12-h light/dark cycle (07:00 lights on - 19:00 lights off) with ad libitum access to food and water, as well as cat’s milk (Whiskas, Mars Canada Inc., Bolton, ON, Canada). Mice were randomly assigned to the groups described.
+
+
+Surgical Procedures
+
We established a well-validated unilateral 6-OHDA mediated Parkinsonian mouse model (Thiele et al., 2012) (Figure 1, Movie S4). 30 minutes before stereotaxic microinjections, mice were intraperitoneally injected with desipramine hydrochloride (2.5 mg/ml, Sigma-Aldrich) and pargyline hydrochloride (0.5 mg/ml, Sigma-Aldrich) at 10 ml/kg (0.9% sterile saline, pH 7.4) to enhance selectivity and efficacy of 6-OHDA induced lesions (Thiele et al., 2012). All surgical procedures were performed using aseptic techniques, and mice were anesthetized using isoflurane (1 - 2%) delivered by 0.4 L/min of medical-grade oxygen (Vitalair 1072, 100% oxygen).
+
Mice were stabilized on a stereotaxic apparatus. Small craniotomies were made above the medial forebrain bundle (MFB) and the A13 nucleus within one randomly assigned hemisphere. Stereotaxic microinjections were performed using a glass capillary (Drummond Scientific, PA, USA; Puller Narishige, diameter 15 – 20 mm) and a Nanoject II apparatus (Drummond Scientific, PA, USA). 240 nL of 6-OHDA (3.6 µg, 15.0 mg/mL; Tocris, USA) was microinjected into the MFB (AP −1.2 mm from bregma; ML ±1.1 mm; DV −5.0 mm from the dura). Sham mice received a vehicle solution (240 nL of 0.2% ascorbic acid in 0.9% saline; Tocris, USA).
+
+Whole Brain Experiments
+
For tracing purposes, a 50:50 mix of AAV8-CamKII-mCherry (Neurophotonics, Laval University, Quebec City, Canada, Lot #820, titre 2×1013 GC/ml) and AAVrg-CAG-GFP (Addgene, Watertown, MA, Catalogue #37825, Lot #V9234, titre ≥ 7×10¹² vg/mL) was injected ipsilateral to 6-OHDA injections at the A13 nucleus in all mice (AP −1.22 mm from bregma; ML ±0.4 mm; DV −4.5 mm from the dura, the total volume of 110 nL at a rate of 23 nl/sec). Post-surgery care was the same for both sham and 6-OHDA injected mice. The animals were sacrificed 29 days after surgery.
+
+
+Photoactivation Experiments
+
36.8 nL of AAVDJ-CaMKIIα-hChR2(H134R)-eYFP (UNC Stanford Viral Gene Core; Stanford, CA, US, Catalogue #AAV36; Lots #3081 and #6878, titres 1.9×1013 and 1.7×1013 GC/mL, respectively) or eYFP control virus (AAVDJ-CaMKIIα-eYFP; Lots #2958 & #5510, titres 7.64×1013 and 2.88×1013 GC/mL, respectively) was injected into the A13’s stereotaxic coordinates (Sharma et al., 2018). A mono-fibre cannula (Doric Lenses, Quebec, Canada, Catalogue #B280-2401-5, MFC_200/230-0.48_5mm_MF2.5_FLT) was implanted slowly 300 μm above the viral injection site. Metabond® Quick Adhesive Cement System (C&B, Parkell, Brentwood, NY, US) and Dentsply Repair Material (Dentsply International Inc., York, PA, USA) were used to fix the ferrule in place. Animals recovered from the viral surgery for 19 days before follow-up behavioral testing. Figure 1 illustrates the timeline of the behavioral tests.
+
+
+
+ChR2 photoactivation
+
A Laserglow Technologies 473 nm laser and driver (LRS-0473-GFM-00100-05, North York, ON, Canada) were used to generate the photoactivation for experiments. The laser was triggered with TTL pulses from either an A.M.P.I. Master-8 stimulator (Jerusalem, Israel) or an Open Ephys PulsePal (Sanworks, Rochester, NY, US) set to 20 Hz with 10-ms pulse width. All fibre optic implants were tested for laser power before implantation (Thorlabs, Saint-Laurent, QC, Canada; optical power sensor (S130C) and meter (PM100D)). The Stanford Optogenetics irradiance calculator was used to estimate the laser power for stimulation (“Stanford Optogenetics Resource Center,” n.d.). A 1×2 fibre-optic rotary joint (Doric Lenses, Quebec, Canada; FRJ_1x2i_FC-2FC_0.22) was used. The animals’ behaviors were recorded with an overhead camera (SuperCircuits, Austin, TX, US; FRJ_1x2i_FC-2FC_0.22; 720 x 480 resolution; 30 fps). The video was processed online (Cleversys, Reston, VA; TopScan V3.0) with a TTL signal output from a National Instruments 24-line digital I/O box (NI, Austin, TX, US; USB-6501) to the Master-8 stimulator.
+
+
+Behavioral Testing
+
+Open Field Test
+
Each mouse was placed in a square arena measuring 70 (W) x 70 (L) x 50 (H) cm with opaque walls and recorded for 30 minutes using a vertically mounted video camera (Model PC165DNR, Supercircuits, Austin, TX, USA; 30 fps). 19 days following surgery, mice were habituated to the open field test (OFT) arena with a patch cable attached for three days in 30 minute sessions to bring animals to a common baseline of activity. On experimental days, after animals were placed in the OFT, a one-minute-on-three-minutes-off paradigm was repeated five times following an initial ten minutes baseline activity. Locomotion was registered when mice travelled a minimum distance of 10 cm at 6 cm/s for 20 frames over a 30-frame segment. When the mouse velocity dropped below 6 cm/s for 20 frames, locomotion was recorded as ending. Bouts of locomotion relate to the number of episodes where the animal met these criteria. Velocity data were obtained from the frame-by-frame results and further processed in a custom Python script to detect instantaneous speeds greater than 2 cm/s (Masini and Kiehn, 2022). All animals that had validated targeting of the A13 region were included in the OFT data presented in the results section, except for one sham ChR2 animal, which showed grooming rather than the typical locomotor phenotypes.
+
+
+Pole Test
+
Mice were placed on a vertical wooden pole (50 cm tall and 1 cm diameter) facing upwards and then allowed to descend the pole into their home cage (Glajch et al., 2012). Animals were trained for three days and tested 2-5 days pre-surgery. Animals were acclimatized 21-22 days post-surgery under two conditions: without a patch cable and with the patch cable attached without photoactivation. On days 24-27, experimental trials were recorded with photoactivation. Video data were recorded for a minimum of three trials (Canon, Brampton, ON, Canada; Vixia HF R52; 1920 x 1080 resolution; 60 fps). A blinded scorer recorded the times for the following events: the hand release of the animal’s tail, the animal fully turning to descend the pole, and the animal reaching the base of the apparatus. Additionally, partial falls, where the animal slipped down the pole but did not reach the base, and full falls, where the animal fell to the base, were recorded separately. All validated animals were included in the quantified data, including the sham ChR2 animal that began grooming in the OFT upon photoactivation. This animal displayed proficiency in performing the PT during photoactivation. It started grooming upon completion of the task when photoactivation was on. One sham ChR2 animal was photostimulated at 1 mW since it would jump off the apparatus at higher stimulation intensities.
+
+
+
+Immunohistochemistry
+
+A13 and SNc region
+
Post hoc analysis of the tissue was performed to confirm the 6-OHDA lesion and validate the targeting of the A13 region. Following behavioral testing, animals underwent a photoactivation protocol to activate neurons below the fibre optic tip (Koblinger et al., 2018). Animals were placed in an OFT for ten minutes before receiving three minutes of photoactivation. Ten minutes later, the animals were returned to their home cage. 90 minutes post photoactivation, animals were deeply anaesthetised with isoflurane and then transcardially perfused with room temperature PBS followed by cold 4% paraformaldehyde (PFA) (Sigma-Aldrich, Catalogue #441244-1KG). The animals were decapitated, and the whole heads were incubated overnight in 4% PFA at 4°C before the fibre optic was removed and the brain removed from the skull. The brain tissue was post-fixed for another 6 - 12 hours in 4% PFA at 4°C then transferred to 30% sucrose solution for 48 - 72 hours. The tissue was embedded in VWR® Clear Frozen Section Compound (VWR International LLC, Radnor, PA, US) and sectioned coronally at 40 or 50 μm using a Leica cryostat set to −21°C (CM 1850 UV, Concord, ON, Canada). Sections from the A13 region (−0.2 to −2.0 mm past bregma) and the SNc (−2.2 to −4.0 past bregma) were collected and stored in PBS containing 0.02% (w/v) sodium azide (EM Science, Catalogue #SX0299-1, Cherry Hill, NJ, US) (Keith B. J. Franklin and Paxinos, 2008).
+
Immunohistochemistry staining was done on free-floating sections. The A13 sections were labelled for c-Fos, TH, and GFP (to enhance eYFP viral signal), and received a DAPI stain to identify nuclei. The SNc sections were stained with TH and DAPI (Table 1). Sections were washed in PBS (3 x 10 mins) then incubated in a blocking solution comprised of PBS containing 0.5% Triton X-100 (Sigma-Aldrich, Catalogue #X100-500ML, St. Louis, MO, US) and 5% donkey serum (EMD Millipore, Catalogue #S30-100ML, Billerica, MA, USA) for 1 hour. This was followed by overnight (for SNc sections) or 24-hour (for A13 sections) incubation in a 5% donkey serum PBS primary solution at room temperature. On day 2, the tissue was washed in PBS (3 x 10 mins) before being incubated in a PBS secondary solution containing 5% donkey serum for 2 hours (for SNc tissue) or 4 hours (for A13 tissue). The secondary was washed with a PBS solution containing 1:1000 DAPI for 10 mins, followed by a final set of PBS washes (3 x 10 mins). Tissue was mounted on Superfrost® micro slides (VWR, slides, Radnor, PA, US) with mounting media (Vectashield®, Vector Laboratories Inc., Burlingame, CA, US), covered with #1 coverslips (VWR, Radnor, PA, US) then sealed.
+
+
+
List of antibodies used for immunohistochemical staining of the A13 and SNc regions, as well as the whole brain.
+
+
+
+
+Whole Brain
+
Mice were deeply anesthetized with isoflurane and transcardially perfused with PBS, followed by 4% PFA. To prepare for whole brain imaging, brains were first extracted and postfixed overnight in 4% PFA at 4°C. The next day, a modified iDISCO method (Renier et al., 2014) was used to clear the samples and perform quadruple immunohistochemistry in whole brains. The modifications include prolonged incubation and the addition of SDS for optimal labelling. The antibodies used are listed in Table 1 and the protocol is provided in Table 2.
All tissue was initially scanned with an Olympus VS120-L100 Virtual Slide Microscope (UPlanSApo, 10x and 20x, NA = 0.4 and 0.75). Standard excitation and emission filter cube sets were used (DAPI, FITC, TRITC, Cy5), and images were acquired using an Orca Flash 4.0 sCMOS monochrome camera (Hamamatsu, Bridgewater Township, NJ, US). For c-Fos immunofluorescence, A13 sections of the tissue were imaged with a Leica SP8 FALCON (FAst Lifetime CONtrast) scanning confocal microscope equipped with a tunable laser and using a 63x objective (HC PlanApo, NA = 1.40).
+
SNc images were imported into Adobe Illustrator, where the SNc (Fougère et al., 2021), including the pars lateralis (SNl), was delineated using the TH immunostaining together with the medial lemniscus and cerebral peduncle as landmarks (bregma −3.09 and −3.68) (Iancu et al., 2005; Keith B. J. Franklin and Paxinos, 2008; Stott and Barker, 2014). Cell counts were obtained using a semi-automated approach using an Ilastik (v1.4.0b15) (Berg et al., 2019) trained model followed by corrections by a blinded counter (Fougère et al., 2021; Iancu et al., 2005). Targeting was confirmed on the 10x overview scans of the A13 region tissue by the presence of eYFP localized in the mZI around the A13 TH+ nucleus, the fibre optic tip being visible near the mZI and A13 nucleus, and the presence of c-Fos positive cells in ChR2+ tissue. C-Fos expression colocalization within the A13 region was performed using confocal images. The mZI & A13 region was identified with the 3rd ventricle and TH expression as markers (Keith B. J. Franklin and Paxinos, 2008).
+
+
+Whole Brain Experiments
+
Cleared whole brain samples were imaged using a light-sheet microscope (LaVision Biotech UltraMicroscope, LaVision, Bielefeld, Germany) with an Olympus MVPLAPO 2x objective with 4x optical zoom (NA = 0.475) and a 5.7 mm dipping cap that is adjusted for the high refractive index of 1.56. The brain samples were imaged in an ethyl cinnamate medium to match the refractive indices and illuminated by three sheets of light bilaterally. Each light sheet was 5 µm thick, and the width was set at 30% to ensure sufficient illumination at the centroid of the sample. Laser power intensities and chromatic aberration corrections used for each laser were as follows: 10% power for 488 nm laser, 5% power for 561 nm laser with 780 nm correction, 40% power for 640 nm laser with 960 nm correction, and 100% power for 785 nm laser with 1,620 nm correction. Each sample was imaged coronally in 8 by 6 squares with 20% overlap (10,202 µm by 5,492 µm in total) and a z-step size of 15 µm (xyz resolution = 0.813 µm x 0.813 µm x 15 µm). While an excellent choice for our work, confocal microscopy offers better resolution at the expense of time. To gain a better resolution using a light-sheet microscope in select regions (eg. SNc and A13 cells), we increased the optical zoom to 6.3x.
+
+
+A13 Connectome Analysis
+
Images were processed using ImageJ software (Schneider et al., 2012). Raw images were stitched, and a z-encoded maximum intensity projection across a 90 µm thick optical section was obtained across each brain. 90 µm sections were chosen because the 2008 Allen reference atlas images are spaced out at around 100 µm. Brains with insufficient quality in labeling were excluded from analysis (n = 1 of three sham and n = 3 of six 6-OHDA mice). Instructions for identifying YFP+ or TH+ cells to annotate were provided to the manual counters. YFP+ and TH+ cells were manually counted using the Cell Counter Plug-In (ImageJ). mCherry+ fibers were segmented semi-automatically using Ilastik software (Berg et al., 2019) and quantified using particle analysis in ImageJ. Images and segmentations were imported into WholeBrain software to be registered with the 2008 Allen reference atlas (Fürth et al., 2018). The TO-PRO™-3 and TH channels were used as reference channels to register each section to a corresponding atlas image. ImageJ quantifications of cell and fiber segmentations were exported in XML formats and registered using WholeBrain software. To minimize the influence of experimental variation on the total labeling of neurons and fibers, the afferent cell counts or efferent fiber areas in each brain region were column divided by the total number found in a brain to obtain the proportion of total inputs and outputs. Connectome analyses were performed using custom R scripts (L. H. Kim et al., 2021). For interregional correlation analyses, the data were normalized to a log10 value to reduce variability and bring brain regions with high and low proportions of cells and fibers to a similar scale. The consistency of afferent and efferent proportions between mice was compared in a pairwise manner using Spearman’s correlation (Figure S5).
+
+
+Quantification of 6-OHDA mediated TH<sup>+</sup> cell loss
+
The percentage of TH+ cell loss was quantified to confirm 6-OHDA mediated SNc lesions. TH+ cells within ZI, VTA and SNc areas from 90 µm thick optical brain slice images (AP: −0.655 to −3.88 mm from bregma) were manually counted by two blinded counters (n = 3 sham and n = 6 6-OHDA mice; ZI region in 2 of 6 6-OHDA mice were excluded due to presence of abnormal scarring/healing at the injection site of viruses). Subsequently, WholeBrain software (Fürth et al., 2018) was used to register and tabulate TH+ cells in the contralesional and ipsilesional brain regions of interest. Counts obtained from the two counters were averaged per region. The percentage of TH+ cell loss was calculated by dividing the difference in counts between contralesional and ipsilesional sides by the contralesional side count and multiplying by 100%.
+
+
+
+Statistical analyses
+
All data were tested for normality using a Shapiro-Wilk test to determine the most appropriate statistical tests. The percent ipsilesional TH+ neuron loss within the SNc as defined above using a Pearson correlation (Fougère et al., 2021) was used to ascertain the effect of the 6-OHDA lesion on behavior. A Wilcoxon rank-sum test was performed for comparisons within subjects at two timepoints where normality failed, and the central limit theorem could not be applied. The two groups were compared using an unpaired t-test with Welch’s correction. A mixed model ANOVA (MM ANOVA) was used to compare the effects of group type, injection type and time. Additionally, Mauchly’s test of sphericity was performed to account for differences in variability within the repeated measures design. A Greenhouse Geisser correction was applied to all ANOVAs where Mauchly’s test was significant for RM and MM ANOVAs. The post hoc multiple comparisons were run when the respective ANOVAs reached significance using Dunnett’s or Dunn’s tests for repeated measures of parametric and non-parametric tests, respectively. The pre-stimulation timepoints were used as the control time point to determine if stimulation altered behavior. A Bonferroni correction was added for post hoc comparisons following a MM ANOVA between groups at given time points to control for alpha value inflation. All correlations, t-tests, and ANOVAs were performed, and graphs were created using Prism version 9.3.1 (Graphpad) or SPSS (IBM, 28.0.1.0). Full statistical reporting is in Supplemental Statistics.xls.
+
+
+Figures
+
Figures were constructed using Adobe Photoshop, Illustrator, and Biorender.
+
+
+
+
+
+Author Contributions
+
LHK and AL performed experiments and prepared figures. MAT and PJW edited figures. LHK, AL, ZHT and PJW conceived and designed the research and interpreted the results. PJW procured funding for the experiments. SS and AL performed surgeries for lesions, optogenetic experiments and conducted behavioral experiments. LHK and SEAE optimized light-sheet imaging. MAT, ST, and CM performed manual cell counting. TC performed analysis and prepared figures on gait analysis. LHK, AL, TC, ZHT, and PJW drafted the manuscript. All authors reviewed and approved the final version of the manuscript.
+
+
+Competing Interest Statement
+
None.
+
+
+ACKNOWLEDGEMENTS
+
We would like to acknowledge support from Whelan and Kiss Labs and technical support from Hotchkiss Brain Institute Advanced Microscopy Platform Core Facility, Cumming School of Medicine Optogenetics Platform Core Facility and Drs. David Elliot, Jonathan Epp, Young Ou, and Lothar Resch. We acknowledge studentships from Parkinson Alberta (LHK), Parkinson Canada (LHK), Canadian Open Neuroscience Platform (AL), Cumming School of Medicine (AL, LHK), Faculty of Graduate Studies (AL, LHK), and the Faculty of Veterinary Medicine (CM, ST). This research was supported by grants to PJW provided by a Canadian Institutes of Health Research Project Grant (PJT-173511), Wings for Life, NSERC (RGPIN/04394-2019) as well as ZHTK from NSERC (RPGIN/04126-2017).
+
+
+DATA AVAILABILITY
+
All datasets and code will be made available on a public repository.
+
+
+REFERENCES
+BergS, KutraD, KroegerT, StraehleCN, KauslerBX, HauboldC, SchieggM, AlesJ, BeierT, RudyM, ErenK, CervantesJI, XuB, BeuttenmuellerF, WolnyA, ZhangC, KoetheU, HamprechtFA, KreshukA. 2019. ilastik: interactive machine learning for (bio)image analysis. Nat Methods16:1226–1232.
+BlomstedtP, Stenmark PerssonR, HarizG-M, LinderJ, FredricksA, HäggströmB, PhilipssonJ, ForsgrenL, HarizM. 2018. Deep brain stimulation in the caudal zona incerta versus best medical treatment in patients with Parkinson’s disease: a randomised blinded evaluation. J Neurol Neurosurg Psychiatry89:710–716.
+BoixJ, PadelT, PaulG. 2015. A partial lesion model of Parkinson’s disease in mice--characterization of a 6-OHDA-induced medial forebrain bundle lesion. Behav Brain Res284:196–206.
+BoltonAD, MurataY, KirchnerR, KimS-Y, YoungA, DangT, YanagawaY, Constantine-PatonM. 2015. A Diencephalic Dopamine Source Provides Input to the Superior Colliculus, where D1 and D2 Receptors Segregate to Distinct Functional Zones. Cell Rep13:1003–1015.
+BordiaT, PerezXA, HeissJ, ZhangD, QuikM. 2016. Optogenetic activation of striatal cholinergic interneurons regulates L-dopa-induced dyskinesias. Neurobiol Dis91:47–58.
+BossonA, BoisseauS, BuissonA, SavastaM, AlbrieuxM. 2015. Disruption of dopaminergic transmission remodels tripartite synapse morphology and astrocytic calcium activity within substantia nigra pars reticulata. Glia63:673–683.
+BraakH, Del TrediciK, RübU, de VosRAI, Jansen SteurENH, BraakE.2003. Staging of brain pathology related to sporadic Parkinson’s disease. Neurobiol Aging24:197–211.
+CaireF, RanouxD, GuehlD, BurbaudP, CunyE. 2013. A systematic review of studies on anatomical position of electrode contacts used for chronic subthalamic stimulation in Parkinson’s disease. Acta Neurochir155:1647–54; discussion 1654.
+CareyRJ. 1991. Chronic L-dopa treatment in the unilateral 6-OHDA rat: evidence for behavioral sensitization and biochemical tolerance. Brain Res568:205–214.
+CenciMA, BjörklundA. 2020. Animal models for preclinical Parkinson’s research: An update and critical appraisal. Prog Brain Res252:27–59.
+ChenF, QianJ, CaoZ, LiA, CuiJ, ShiL, XieJ.2023. Chemogenetic and optogenetic stimulation of zona incerta GABAergic neurons ameliorates motor impairment in Parkinson’s disease. iScience26:107149.
+ChoiEA, McNallyGP. 2017. Paraventricular Thalamus Balances Danger and Reward. J Neurosci37:3018–3029.
+CohenJ.1988. Statistical power analysis for the behavioral sciences New York. NY: Academic54.
+CostallB, NaylorRJ, PycockC. 1976. Non-specific supersensitivity of striatal dopamine receptors after 6-hydroxydopamine lesion of the nigrostriatal pathway. Eur J Pharmacol35:276–283.
+DauerW, PrzedborskiS. 2003. Parkinson’s disease: mechanisms and models. Neuron39:889–909.
+EatonMJ, WagnerCK, MooreKE, LookinglandKJ. 1994. Neurochemical identification of A13 dopaminergic neuronal projections from the medial zona incerta to the horizontal limb of the diagonal band of Broca and the central nucleus of the amygdala. Brain Res659:201–207.
+FerrayeMU, DebûB, FraixV, GoetzL, ArdouinC, YelnikJ, Henry-LagrangeC, SeigneuretE, PiallatB, KrackP, Le BasJ-F, BenabidA-L, ChabardèsS, PollakP.2010. Effects of pedunculopontine nucleus area stimulation on gait disorders in Parkinson’s disease. Brain133:205–214.
+FougèreM, van der ZouwenCI, BoutinJ, NeszvecskoK, SarretP, RyczkoD.2021. Optogenetic stimulation of glutamatergic neurons in the cuneiform nucleus controls locomotion in a mouse model of Parkinson’s disease. Proc Natl Acad Sci U S A118. doi:10.1073/pnas.2110934118
+FürthD, VaissièreT, TzortziO, XuanY, MärtinA, LazaridisI, SpigolonG, FisoneG, TomerR, DeisserothK, CarlénM, MillerCA, RumbaughG, MeletisK. 2018. An interactive framework for whole-brain maps at cellular resolution. Nat Neurosci21:139–149.
+GlajchKE, FlemingSM, SurmeierDJ, OstenP. 2012. Sensorimotor assessment of the unilateral 6-hydroxydopamine mouse model of Parkinson’s disease. Behav Brain Res230:309–316.
+GlicksteinM, SteinJ. 1991. Paradoxical movement in Parkinson’s disease. Trends Neurosci14:480–482.
+GoldowitzD. 2010. Allen Reference Atlas. A Digital Color Brain Atlas of the C57BL/6J Male Mouse - by H. W. Dong. Genes, Brain and Behavior. doi:10.1111/j.1601-183x.2009.00552.x
+GradinaruV, MogriM, ThompsonKR, HendersonJM, DeisserothK. 2009. Optical deconstruction of parkinsonian neural circuitry. Science324:354–359.
+GutNK, WinnP. 2015. Deep Brain Stimulation of Different Pedunculopontine Targets in a Novel Rodent Model of Parkinsonism. Journal of Neuroscience. doi:10.1523/jneurosci.3646-14.2015
+HaeryL, DevermanBE, MathoKS, CetinA, WoodardK, CepkoC, GuerinKI, RegoMA, ErsingI, BachleSM, KamensJ, FanM. 2019. Adeno-Associated Virus Technologies and Methods for Targeted Neuronal Manipulation. Front Neuroanat13:93.
+HamaniC, MoroE, LozanoAM. 2011. The pedunculopontine nucleus as a target for deep brain stimulation. J Neural Transm118:1461–1468.
+HansenHH, RoostaluU, Hecksher-SørensenJ. 2020. Whole-brain three-dimensional imaging for quantification of drug targets and treatment effects in mouse models of neurodegenerative diseases. Neural Regeneration Res15:2255–2257.
+HoffmanBJ, PalkovitsM, PaiakK, HamsonSR, MezeyÉ. 1997. Regulation of Dopamine Transporter mRNA Levels in the Central Nervous System In: Goldstein DS, Eisenhofer G, McCarty R, editors. Advances in Pharmacology. Academic Press. pp. 202–206.
+HormigoS, ZhouJ, Castro-AlamancosMA. 2020. Zona Incerta GABAergic Output Controls a Signaled Locomotor Action in the Midbrain Tegmentum. eNeuro7. doi:10.1523/ENEURO.0390-19.2020
+HughesAJ, Ben-ShlomoY, DanielSE, LeesAJ. 1992. What features improve the accuracy of clinical diagnosis in Parkinson’s disease: a clinicopathologic study. Neurology42:1142–1146.
+IancuR, MohapelP, BrundinP, PaulG. 2005. Behavioral characterization of a unilateral 6-OHDA-lesion model of Parkinson’s disease in mice. Behav Brain Res162:1–10.
+JagmagSA, TripathiN, ShuklaSD, MaitiS, KhuranaS. 2015. Evaluation of Models of Parkinson’s Disease. Front Neurosci9:503.
+KeithB. J.FranklinMA, PaxinosG. 2008. The Mouse Brain in Stereotaxic Coordinates, Compact: The Coronal Plates and Diagrams. Elsevier Science.
+KimbroughA, LurieDJ, CollazoA, KreifeldtM, SidhuH, MacedoGC, D’EspositoM, ContetC, GeorgeO. 2020. Brain-wide functional architecture remodeling by alcohol dependence and abstinence. Proc Natl Acad Sci U S A117:2149–2159.
+KimL, ChomiakT, TranMA, TamS, McPhersonC, EatonSEA, OuY, KissZHT, WhelanPJ.2021. Global remodelling of afferent and efferent projections of the A13 region following unilateral nigrostriatal degeneration using 6-hydroxydopamineNeuroscience Meeting Planner. Presented at the Society for Neuroscience.
+KimLH, ChomiakT, TranM, TamS, McPhersonC, EatonSEA, OuY, ReschL, KissZHT, WhelanPJ.2021. Substantia nigra degradation results in widespread changes in medial zona incerta afferent and efferent connectomics.
+KimLH, SharmaS, SharplesSA, MayrKA, KwokCHT, WhelanPJ. 2017. Integration of Descending Command Systems for the Generation of Context-Specific Locomotor Behaviors. Front Neurosci11:581.
+KishSJ, TongJ, HornykiewiczO, RajputA, ChangL-J, GuttmanM, FurukawaY. 2008. Preferential loss of serotonin markers in caudate versus putamen in Parkinson’s disease. Brain131:120–131.
+KoblingerK, Jean-XavierC, SharmaS, FüzesiT, YoungL, EatonSEA, KwokCHT, BainsJS, WhelanPJ. 2018. Optogenetic Activation of A11 Region Increases Motor Activity. Front Neural Circuits12:86.
+KonnovaEA, Translational NeurogeneticsUnit, Wallenberg NeuroscienceCenter, Lund University, Lund, Sweden, Swanberg M, Translational NeurogeneticsUnit, Wallenberg NeuroscienceCenter, Lund University, Lund, Sweden. 2018. Animal models of Parkinson’s diseaseParkinson’s Disease: Pathogenesis and Clinical Aspects. Codon Publications. pp. 83–106.
+KravitzAV, FreezeBS, ParkerPRL, KayK, ThwinMT, DeisserothK, KreitzerAC. 2010. Regulation of parkinsonian motor behaviours by optogenetic control of basal ganglia circuitry. Nature466:622–626.
+LaneEL, CheethamSC, JennerP. 2006. Does contraversive circling in the 6-OHDA-lesioned rat indicate an ability to induce motor complications as well as therapeutic effects in Parkinson’s disease?Exp Neurol197:284–290.
+LeeCS, SchulzerM, MakE, HammerstadJP, CalneS, CalneDB. 1995. Patterns of asymmetry do not change over the course of idiopathic parkinsonism: implications for pathogenesis. Neurology45:435–439.
+LiL-X, LiY-L, WuJ-T, SongJ-Z, LiX-M. 2022. Glutamatergic Neurons in the Caudal Zona Incerta Regulate Parkinsonian Motor Symptoms in Mice. Neurosci Bull38:1–15.
+LimS-Y, FoxSH, LangAE. 2009. Overview of the extranigral aspects of Parkinson disease. Arch Neurol66:167–172.
+LiS, ShiY, KirouacGJ. 2014. The hypothalamus and periaqueductal gray are the sources of dopamine fibers in the paraventricular nucleus of the thalamus in the rat8:136.
+LizarragaKJ, JagidJR, LucaCC. 2016. Comparative effects of unilateral and bilateral subthalamic nucleus deep brain stimulation on gait kinematics in Parkinson’s disease: a randomized, blinded study. J Neurol263:1652–1656.
+LiZ, RizziG, TanKR. 2021. Zona incerta subpopulations differentially encode and modulate anxiety. Sci Adv7:eabf6709.
+MagnoLAV, Tenza-FerrerH, CollodettiM, AguiarMFG, Rodrigues APC, da Silva RS, Silva J do P, Nicolau NF, Rosa DVF, Birbrair A, Miranda DM, Romano-Silva MA. 2019. Optogenetic Stimulation of the M2 Cortex Reverts Motor Dysfunction in a Mouse Model of Parkinson’s Disease. J Neurosci39:3234–3248.
+MasiniD, KiehnO. 2022. Targeted activation of midbrain neurons restores locomotor function in mouse models of parkinsonism. Nat Commun13:504.
+MatsuuraK, KabutoH, MakinoH, OgawaN. 1997. Pole test is a useful method for evaluating the mouse movement disorder caused by striatal dopamine depletion. J Neurosci Methods73:45–48.
+MatzukMM, SaperCB. 1985. Preservation of hypothalamic dopaminergic neurons in Parkinson’s disease. Ann Neurol18:552–555.
+MeredithGE, RademacherDJ. 2011. MPTP mouse models of Parkinson’s disease: an update. J Parkinsons Dis1:19–33.
+MessanviF, Eggens-MeijerE, RoozendaalB, van der WantJJ.2013. A discrete dopaminergic projection from the incertohypothalamic A13 cell group to the dorsolateral periaqueductal gray in rat. Front Neuroanat7:41.
+MitrofanisJ. 2005. Some certainty for the “zone of uncertainty”? Exploring the function of the zona incerta. Neuroscience130:1–15.
+MitrofanisJ, MikuleticL. 1999. Organisation of the cortical projection to the zona incerta of the thalamus. J Comp Neurol412:173–185.
+MokD, MogensonGJ. 1986. Contribution of zona incerta to osmotically induced drinking in rats. Am J Physiol251:R823–32.
+MoonHC, WonSY, KimEG, KimHK, ChoCB, ParkYS. 2018. Effect of optogenetic modulation on entopeduncular input affects thalamic discharge and behavior in an AAV2-α-synuclein-induced hemiparkinson rat model. Neurosci Lett662:129–135.
+MoriyaS, YamashitaA, MasukawaD, SetoyamaH, HwangY, YamanakaA, KuwakiT. 2020. Involvement of A13 dopaminergic neurons located in the zona incerta in nociceptive processing: a fiber photometry study. Mol Brain13:60.
+MoroE, HamaniC, PoonY-Y, Al-KhairallahT, DostrovskyJO, HutchisonWD, LozanoAM. 2010. Unilateral pedunculopontine stimulation improves falls in Parkinson’s disease. Brain133:215–224.
+NegishiK, PayantMA, SchumackerKS, WittmannG, ButlerRM, LechanRM, SteinbuschHWM, KhanAM, CheeMJ. 2020. Distributions of hypothalamic neuron populations coexpressing tyrosine hydroxylase and the vesicular GABA transporter in the mouse. J Comp Neurol528:1833–1855.
+NogaBR, WhelanPJ. 2022. The Mesencephalic Locomotor Region: Beyond Locomotor Control. Front Neural Circuits16:884785.
+NonnekesJ, BereauM, BloemBR. 2020. Freezing of Gait and Its Levodopa Paradox. JAMA Neurol77:287–288.
+NonnekesJ, SnijdersAH, NuttJG, DeuschlG, GiladiN, BloemBR. 2015. Freezing of gait: a practical approach to management. Lancet Neurol14:768–778.
+OgawaN, HiroseY, OharaS, OnoT, WatanabeY. 1985. A simple quantitative bradykinesia test in MPTP-treated mice. Res Commun Chem Pathol Pharmacol50:435–441.
+OgundeleOM, LeeCC, FrancisJ. 2017. Thalamic dopaminergic neurons projects to the paraventricular nucleus-rostral ventrolateral medulla/C1 neural circuit. The Anatomical Record. doi:10.1002/ar.23528
+OkunMS, FooteKD. 2010. Parkinson’s disease DBS: what, when, who and why? The time has come to tailor DBS targets. Expert Rev Neurother10:1847–1857.
+OssowskaK. 2019. Zona incerta as a therapeutic target in Parkinson’s disease. Journal of Neurology. doi:10.1007/s00415-019-09486-8
+PeoplesC, SpanaS, AshkanK, BenabidA-L, StoneJ, BakerGE, MitrofanisJ. 2012. Photobiomodulation enhances nigral dopaminergic cell survival in a chronic MPTP mouse model of Parkinson’s disease. Parkinsonism Relat Disord18:469–476.
+Perez-LloretS, BarrantesFJ. 2016. Deficits in cholinergic neurotransmission and their clinical correlates in Parkinson’s disease. NPJ Parkinsons Dis2:16001.
+PérierC, TremblayL, FégerJ, HirschEC. 2002. Behavioral consequences of bicuculline injection in the subthalamic nucleus and the zona incerta in rat. J Neurosci22:8711–8719.
+PérierC, VilaM, FégerJ, AgidY, HirschEC. 2000. Functional activity of zona incerta neurons is altered after nigrostriatal denervation in hemiparkinsonian rats. Exp Neurol162:215–224.
+PerovićM, MladenovićA, RakićL, RuzdijićS, KanazirS. 2005. Increase of GAP-43 in the rat cerebellum following unilateral striatal 6-OHDA lesion. Synapse56:170–174.
+PlahaP, KhanS, GillSS. 2008. Bilateral stimulation of the caudal zona incerta nucleus for tremor control. J Neurol Neurosurg Psychiatry79:504–513.
+PowerBD, MitrofanisJ. 1999. Evidence for extensive inter-connections within the zona incerta in rats. Neurosci Lett267:9–12.
+RenierN, WuZ, SimonDJ, YangJ, ArielP, Tessier-LavigneM. 2014. iDISCO: a simple, rapid method to immunolabel large tissue samples for volume imaging. Cell159:896–910.
+RequejoC, López-de-IpiñaK, Ruiz-OrtegaJÁ, FernándezE, CalvoPM, Morera-HerrerasT, MiguelezC, Cardona-GrifollL, CepedaH, UgedoL, LafuenteJV. 2020. Changes in Day/Night Activity in the 6-OHDA-Induced Experimental Model of Parkinson’s Disease: Exploring Prodromal Biomarkers. Front Neurosci14:590029.
+RoostaluU, SalinasCBG, ThorbekDD, SkytteJL, FabriciusK, BarkholtP, JohnLM, JurtzVI, KnudsenLB, JelsingJ, VrangN, HansenHH, Hecksher-SørensenJ. 2019. Quantitative whole-brain 3D imaging of tyrosine hydroxylase-labeled neuron architecture in the mouse MPTP model of Parkinson’s disease. Dis Model Mech12. doi:10.1242/dmm.042200
+RyanMB, Bair-MarshallC, NelsonAB. 2018. Aberrant Striatal Activity in Parkinsonism and Levodopa-Induced Dyskinesia. Cell Rep23:3438–3446.e5.
+RyczkoD, ConeJJ, AlpertMH, GoetzL, AuclairF, DubéC, ParentM, RoitmanMF, AlfordS, DubucR. 2016. A descending dopamine pathway conserved from basal vertebrates to mammals. Proc Natl Acad Sci U S A113:E2440–9.
+SahelJ-A, Boulanger-ScemamaE, PagotC, ArleoA, GalluppiF, MartelJN, DelauxA, de Saint AubertJ-B, de MontleauC, GutmanE, AudoI, DuebelJ, PicaudS, DalkaraD, BlouinL, TaielM, RoskaB.2021. Partial recovery of visual function in a blind patient after optogenetic therapy. Nat Med27:1223–1229.
+SandersTH, JaegerD. 2016. Optogenetic stimulation of cortico-subthalamic projections is sufficient to ameliorate bradykinesia in 6-ohda lesioned mice. Neurobiol Dis95:225–237.
+SangheraMK, Anselmo-FranciJ, McCannSM. 1991. Effect of Medial Zona Incerta Lesions on the Ovulatory Surge of Gonadotrophins and Prolactin in the Rat. Neuroendocrinology. doi:10.1159/000125931
+SangheraMK, GradyS, SmithW, WoodwardDJ, PorterJC. 1991. Incertohypothalamic A13 dopamine neurons: effect of gonadal steroids on tyrosine hydroxylase. Neuroendocrinology53:268–275.
+ScattonB, Javoy-AgidF, RouquierL, DuboisB, AgidY. 1983. Reduction of cortical dopamine, noradrenaline, serotonin and their metabolites in Parkinson’s disease. Brain Res275:321–328.
+SchneiderCA, RasbandWS, EliceiriKW. 2012. NIH Image to ImageJ: 25 years of image analysis. Nat Methods9:671–675.
+SchwartingRK, BonatzAE, CareyRJ, HustonJP. 1991. Relationships between indices of behavioral asymmetries and neurochemical changes following mesencephalic 6-hydroxydopamine injections. Brain Res554:46–55.
+SharmaS, KimLH, MayrKA, ElliottDA, WhelanPJ. 2018. Parallel descending dopaminergic connectivity of A13 cells to the brainstem locomotor centers. Sci Rep8:7972.
+ShawVE, SpanaS, AshkanK, BenabidA-L, StoneJ, BakerGE, MitrofanisJ. 2010. Neuroprotection of midbrain dopaminergic cells in MPTP-treated mice after near-infrared light treatment. J Comp Neurol518:25–40.
+SimolaN, MorelliM, CartaAR. 2007. The 6-hydroxydopamine model of Parkinson’s disease. Neurotox Res11:151–167.
+SitaLV, EliasCF, BittencourtJC. 2007. Connectivity pattern suggests that incerto-hypothalamic area belongs to the medial hypothalamic system148:949–969.
+Stanford Optogenetics Resource Center. n.d. https://web.stanford.edu/group/dlab/cgi-bin/graph/chart.php
+StefaniA, LozanoAM, PeppeA, StanzioneP, GalatiS, TropepiD, PierantozziM, BrusaL, ScarnatiE, MazzoneP. 2007. Bilateral deep brain stimulation of the pedunculopontine and subthalamic nuclei in severe Parkinson’s disease. Brain130:1596–1607.
+StottSRW, BarkerRA. 2014. Time course of dopamine neuron loss and glial response in the 6-OHDA striatal mouse model of Parkinson’s disease. Eur J Neurosci39:1042–1056.
+ThevathasanW, DebuB, AzizT, BloemBR, BlahakC, ButsonC, CzerneckiV, FoltynieT, FraixV, GrabliD, JointC, LozanoAM, OkunMS, OstremJ, PaveseN, SchraderC, TaiC-H, KraussJK, MoroE, Movement Disorders Society PPN DBS Working Groupin collaboration with the World Society for Stereotactic and Functional Neurosurgery. 2018. Pedunculopontine nucleus deep brain stimulation in Parkinson’s disease: A clinical review. Mov Disord33:10–20.
+ThieleSL, WarreR, NashJE. 2012. Development of a unilaterally-lesioned 6-OHDA mouse model of Parkinson’s disease. J Vis Exp. doi:10.3791/3234
+TsangE, OrlandiniC, SurekaR, CrevennaAH, PerlasE, PrankerdI, MasferrerME, GrossCT.2021. Induction of flight via midbrain projections to the cuneiform nucleus. bioRxiv. doi:10.1101/2021.12.21.473683
+UngerstedtU. 1971. Striatal dopamine release after amphetamine or nerve degeneration revealed by rotational behaviour. Acta Physiol Scand Suppl367:49–68.
+ValverdeS, VandecasteeleM, PietteC, DerousseauxW, GangarossaG, Aristieta ArbelaizA, TouboulJ, DegosB, VenanceL. 2020. Deep brain stimulation-guided optogenetic rescue of parkinsonian symptoms. Nat Commun11:2388.
+VenkataramanA, HunterSC, DhinojwalaM, GhebrezadikD, GuoJ, InoueK, YoungLJ, DiasBG. 2021. Incerto-thalamic modulation of fear via GABA and dopamine. Neuropsychopharmacology46:1658–1668.
+WangX, ChouX-L, ZhangLI, TaoHW. 2020. Zona Incerta: An Integrative Node for Global Behavioral Modulation. Trends Neurosci43:82–87.
+WardasJ, OssowskaK, WolfarthS. 1988. Evidence for the independent role of GABA synapses of the zona incerta-lateral hypothalamic region in haloperidol-induced catalepsy. Brain Res462:378–382.
+YoonHH, NamM-H, ChoiI, MinJ, JeonSR. 2020. Optogenetic inactivation of the entopeduncular nucleus improves forelimb akinesia in a Parkinson’s disease model. Behav Brain Res386:112551.
+YoonHH, ParkJH, KimYH, MinJ, HwangE, LeeCJ, SuhJ-KF, HwangO, JeonSR. 2014. Optogenetic inactivation of the subthalamic nucleus improves forelimb akinesia in a rat model of Parkinson disease. Neurosurgery74:533–40; discussion 540–1.
+ZetterströmT, Herrera-MarschitzM, UngerstedtU. 1986. Simultaneous measurement of dopamine release and rotational behaviour in 6-hydroxydopamine denervated rats using intracerebral dialysis. Brain Res376:1–7.
+ZhanY, WuH, LiuL, LinJ, ZhangS. 2021. Organic solvent-based tissue clearing techniques and their applications. J Biophotonics14:e202000413.
+ZhaoZ-D, ChenZ, XiangX, HuM, XieH, JiaX, CaiF, CuiY, ChenZ, QianL, LiuJ, ShangC, YangY, NiX, SunW, HuJ, CaoP, LiH, ShenWL. 2019. Zona incerta GABAergic neurons integrate prey-related sensory signals and induce an appetitive drive to promote hunting. Nat Neurosci22:921–932.
+ZweigRM, JankelWR, HedreenJC, MayeuxR, PriceDL. 1989. The pedunculopontine nucleus in Parkinson’s disease. Ann Neurol26:41–46.
+
+
+SUPPLEMENTARY MATERIAL
+
+
+
Time course of open field locomotion distance travelled over 30 minutes.
+
(A-B) 30-minute open field experiment group averages for sham (A) and 6-OHDA (B) animals with photoactivation plotted as 1-minute bins of distance travelled. Blue bars indicate 1-minute trials with photoactivation. (C) Locomotion distance travelled for the six sham ChR2 animals at baseline and at the five pre-timepoints compared using a 1-way RM ANOVA (F5,25 = 0.486, P = .783). Data indicate mean ± SEM bars.
+
+
+
+
+
Characterization of A13 region photoactivation temporal dynamics on locomotion initiation. (A)
+
Percent of trials where there was at least one bout of locomotion. Data are plotted as box and whiskers with the horizontal line through the box indicating the group median, interquartile range indicated by the limits of the box, and group minimum and maximum indicated by the whiskers. (B) The average time for the ChR2 group animals to begin locomotion after the onset of photoactivation. Means plotted with error bars indicating ± SEM. Asterisks indicate significant comparisons using the Wilcoxon signed-rank test: ** P ≤ 0.05.
+
+
+
+
+
Preservation of TH<sup>+</sup> A13 cells in Parkinsonian mouse models.
+
Representative slices of SNc (AP: −3.08 mm, A) and A13 region (AP: −1.355 mm, D) following registration with WholeBrain software 64. Full 3D brain is available (see Movie S4). There was a lack of TH+ SNc cells following 6-OHDA injections at the MFB (A). (B, C) Zoomed sections (90 μm thickness) of red boxes in panel A in left to right order. Meanwhile, TH+ VTA cells were preserved bilaterally. In addition, TH+ A13 cells were present ipsilesional to 6-OHDA injections (D). (E, F) Zoomed sections (90 μm thickness) of red boxes in panel D in left to right order. Scale bars are 50 μm. When calculating the percentage of TH+ cell loss normalized to the intact side, there was a significant interaction between the condition group and brain regions (repeated measures two-way ANOVA with post hoc Bonferroni pairwise, sham: n = 3, 6-OHDA: n = 6) G). 6-OHDA treated mice showed a significantly greater percentage of TH+ cell loss in SNc compared to VTA and A13 region (VTA vs. SNc: P = 0.005; A13 region vs. SNc: P = 0.029). In contrast, sham showed no significant difference in TH+ cell loss across SNc, VTA and A13 region regions (P > 0.05). *P ≤ 0.05, and **P ≤ 0.01. Scale bars are 50 μm unless otherwise indicated.
+
+
+
+
+
Example of the injection core in a sham brain for viral tracers and the rostral and caudal spread to the injection site (A13). Viral tracers (AAV8-CamKII-mCherry and AAVrg-CAG-GFP) were mixed 50:50. Light-sheet images around the injection site were obtained with 2x objective, 6.3x optical zoom, and a z-step size of 2 µm (xyz resolution = 0.477 µm x 0.477 µm x 2 µm). Background filtering (median value of 20 pixels and Gaussian smoothing with a sigma value of 10) was performed in ImageJ software 1 and visualized in IMARIS 9.8 (Belfast, United Kingdom). 2008 Allen reference atlas images were overlaid on top of 90 µm maximum intensity projections taken from IMARIS 9.8 (Belfast, United Kingdom):-1.26 mm (A), −1.36 mm (B), and −1.46 mm (C). Zoomed in sections of each white rectangular area at each coordinate (rows ‘i’) are displayed below for each fluorophore (rows ‘ii’). Scale bars for rows ‘i’ are 200 µm and for rows ‘ii are 100 µm.
+
+
+
+
+
The consistency of afferent and efferent proportions across mice was compared in a pairwise manner. An experimental variation on the total labeling of neurons and fibers was minimized by dividing the afferent cell counts or efferent fiber areas in each brain region by the total number found in a brain to obtain the proportion of total inputs and outputs. Using Spearman’s correlation analysis, we found afferent and efferent proportions across animals to be consistent among each other with an average correlation of 0.91 (SEM = 0.02). M1 = mouse #1, M2 = mouse #2, M3 = mouse #3.
+
+
+
Movie S1. Photoactivation of the A13 region in a 6-OHDA model mouse producing increased locomotion in the OFT (2x speed).
+
Movie S2. Photoactivation of the A13 region in a sham mouse producing increased locomotion in the OFT (2x speed).
+
Movie S3. Photoactivation of the A13 region during the pole test in a 6-OHDA model mouse decreases pole descent time (0.5x speed).
+
Movie S4. Video showing TH staining following whole brain imaging and staining in a 6-OHDA model mouse brain. Focus on TH expression in the A13 and SNc regions.
Diffusional kurtosis imaging (DKI) is a methodology for measuring the extent of non-Gaussian diffusion in biological tissue, which has shown great promise in clinical diagnosis, treatment planning and monitoring of many neurological diseases and disorders. However, robust, fast and accurate estimation of kurtosis from clinically feasible data acquisitions remains a challenge. In this study, we first outline a new accurate approach of estimating mean kurtosis via the sub-diffusion mathematical framework. Crucially, this extension of the conventional DKI overcomes the limitation on the maximum b-value of the latter. Kurtosis and diffusivity can now be simply computed as functions of the sub-diffusion model parameters. Second, we propose a new fast and robust fitting procedure to estimate the sub-diffusion model parameters using two diffusion times without increasing acquisition time as for the conventional DKI. Third, our sub-diffusion based kurtosis mapping method is evaluated using both simulations and the Connectome 1.0 human brain data. Exquisite tissue contrast is achieved even when the diffusion encoded data is collected in only minutes. In summary, our findings suggest robust, fast and accurate estimation of mean kurtosis can be realised within a clinically feasible diffusion weighted magnetic resonance imaging data acquisition time.
+
+
+
+
+
+
+
+
+Introduction
+
Diffusion weighted magnetic resonance imaging (DW-MRI) over a period of more than 30 years has become synonymous with tissue microstructure imaging. Measures of how water diffuses in heterogeneous tissues allow indirect interpretation of changes in tissue microstructure (Le Bihan and Johansen-Berg, 2012). DW-MRI has predominantly been applied in the brain, where properties of white matter connections between brain regions are often studied (Lebel et al., 2019), in addition to mapping tissue microstructural properties (Tournier, 2019). Applications outside of the brain have clinical importance as well, and interest is growing rapidly.
+
Generally, DW-MRI involves the setting of diffusion weightings and direction over which diffusion is measured. While diffusion tensor imaging (DTI) can be performed using a single diffusion weighting, a so-called b-shell, and at least six diffusion encoding directions (Le Bihan et al., 2001), other models tend to require multiple b-shells each having multiple diffusion encoding directions. Diffusional kurtosis imaging (DKI) is a primary example of a multiple b-shell, multiple diffusion encoding direction method (Jensen et al., 2005). DKI is considered as an extension of DTI (Jensen and Helpern, 2010;Hansen et al., 2013; Veraart et al., 2011b), where the diffusion process is assumed to deviate away from standard Brownian motion, and the extent of such deviation is measured via the kurtosis metric. Essentially, the increased sampling achieved via DKI data acquisitions allows more complex models to be applied to data (Van Essen et al., 2013; Shafto et al., 2014), in turn resulting in metrics of increased utility for clinical decision making.
+
Recent clinical benefits of using kurtosis metrics over other DW-MRI derived measures have been demonstrated for grading hepatocellular carcinoma (Li et al., 2022b), prognosing chronic kidney disease (Liu et al., 2021), differentiating parotid gland tumours (Huang et al., 2021a), measuring response to radiotherapy treatment in bone tumour (Guo et al., 2022a) and glioblastoma (Goryawala et al., 2022), identifying tissue abnormalities in temporal lobe epilepsy patients with sleep disorders (Guo et al., 2022b) and brain microstructural changes in mild traumatic brain injury (Wang et al., 2022), monitoring of renal function and interstitial fibrosis (Li et al., 2022a), detecting the invasiveness of bladder cancer into muscle (Li et al., 2022d), aiding management of patients with depression (Maralakunte et al., 2022), delineating acute infarcts with prognostic value (Hu et al., 2022), predicting breast cancer metastasis (Zhou et al., 2022), diagnosing Parkinson’s disease (Li et al., 2022c), amongst others reported prior and not listed here.
+
The routine use of DKI in the clinic has nonetheless lagged due the inability to robustly estimate the kurtosis metric (Veraart et al., 2011a; Tabesh et al., 2010; Kuder et al., 2011; Henriques et al., 2021). A known requirement for estimating kurtosis in DKI is to restrict the maximum b-value to 2000 s/mm2-3000 s/mm2 for brain studies (Jensen et al., 2005; Jensen and Helpern, 2010; Poot et al., 2010), with the optimal maximum b-value found to be dependent on tissue type (Poot et al., 2010). This suggests that the traditional kurtosis model is less accurate at representing the diffusion signal at large b-values. Moreover, multiple b-shell, multiple direction high quality DW-MRI data can take many minutes to acquire, which poses challenges for clinical imaging protocols involving a multitude of MRI contrasts already taking tens of minutes to execute. Reduction of DKI data acquisition times through parallel imaging, optimisation of b-shells and directions have been investigated (Zong et al., 2021; Heidemann et al., 2010; Zelinski et al., 2008), and DW-MRI data necessary for DKI analysis has been shown to supersede the data required for DTI (Veraart et al., 2011b). Therefore, an optimised DKI protocol can potentially replace clinical DTI data acquisitions without adversely affecting the estimation of DTI metrics.
+
For DKI to become a routine clinical tool, DW-MRI data acquisition needs to be fast and provides a robust estimation of kurtosis. The ideal protocol should have a minimum number of b-shells and diffusion encoding directions in each b-shell. The powder averaging over diffusion directions improves the signal-to-noise ratio of the DW-MRI data used for parameter estimation. Whilst this approach loses out on directionality of the kurtosis, it nonetheless provides a robust method of estimating mean kurtosis (Henriques et al., 2021), a metric of significant clinical value.
+
Instead of attempting to improve an existing model-based approach for kurtosis estimation, as has been considered by many others, we considered the problem from a different perspective. In view of the recent generalisation of the various models applicable to DW-MRI data (Yang et al., 2022), the sub-diffusion framework provides new, unexplored opportunities, for fast and robust kurtosis mapping. We report on our investigation into the utility of the sub-diffusion model for practically useful mapping of mean kurtosis.
+
+
+Results
+
Simulation studies were conducted to establish the requirements on the number of diffusion times and the separation between them for accurate estimation of mean kurtosis based on the subdiffusion model augmented with random Gaussian noise, following (10). Testing and validation was performed using the human Connectome 1.0 brain dataset (Tian et al., 2022). The 2 × 2 × 2mm3 resolution data were obtained using two diffusion times (Δ = 19, 49ms) with a pulse duration of δ = 8ms and G = 31,68,105,142,179, 216, 253, 290 mT/m, respectively generating b-values = 50, 350, 800, 1500, 2400, 3450, 4750, 6000 s/mm2, and b-values = 200, 950, 2300, 4250, 6750, 9850, 13500, 17800 s/mm2, according to b-value = (γδG)2(Δ - δ/3). Up to 64 diffusion encoding directions per b-shell were set. The traditional method for mean kurtosis estimation was implemented (producing KDKI), which is limited to the use of DW-MRI generated using a single diffusion time (Jensen et al., 2005; Jensen and Helpern, 2010; Veraart et al., 2011a; Poot et al., 2010), alongside our implementation based on the sub-diffusion model (3), wherein mean kurtosis K* is computed as a function of the sub-diffusion model parameter β (refer to (9)) using either a single or multiple diffusion times.
+
+Multiple diffusion times for robust and accurate mean kurtosis estimation
+
In our simulations we tested up to five distinct diffusion times to generate b-values. Figure 1 illustrates the effects of the number of diffusion times on the parameter estimation at various SNR levels. We draw attention to a number of features in the plots. First, as SNR is increased from 5 to 20 (rows 1-3), the variability in the estimated parameters (Dβ, β, K*) decreases. Second, increasing the number of distinct diffusion times used for parameter estimation decreases estimation variability, with the most significant improvement when increasing from one to two diffusion times (rows 1-3). Third, sampling with two distinct diffusion times provides more robust parameter estimates than sampling twice as many b-values using one diffusion time (compare 2 and 2′ violin plots, rows 1-3). Fourth, the last row (row 4) highlights the improvement in the coefficient of variation (CV) for each parameter estimate with increasing SNR. This result again confirms that the most pronounced decline ofCV occurs when increasing from one to two diffusion times, and parameter estimations can be performed more robustly using DW-MRI data with a relatively high SNR.
+
In Figure 2 we provide simulation results evaluating the choice of the two distinct diffusion times (assuming Δ1 < Δ2) by measuring the goodness-of-fit of the model. The smaller of the two diffusion times is stated along the abscissa, and the difference, i.e., Δ2 - Δ1, is plotted along the ordinate. The quality of fitting was measured using the coefficient of determination (the larger the R2 value, the better the goodness-of-fit of the model) for each combination of abscissa and ordinate values. The conclusion from this figure is that Δ1 should be small, while Δ2 should be as large as practically plausible. Note, DW-MRI echo time was not considered in this simulation, but as Δ2 increases, the echo time has to proportionally increase. Because of the inherent consequence of decreasing SNR with echo time, special attention should be paid to the level of SNR achievable with the use of a specific Δ2. Nonetheless, our findings suggest that when Δ1 = 8 ms, Δ2 can be set as small as 21 ms to achieve an R2 > 0.90 with an SNR as low as 5. If Δ1 is increased past 15 ms, then the separation between Δ1 and Δ2 has to increase as well, and such choices benefit from an increased SNR in the DW-MRI data. The Connectome 1.0 DW-MRI data was obtained using Δ1 = 19 ms and Δ2 = 49 ms, leading to a separation of 30 ms. For this data it is expected that with SNR = 5 an R2 around 0.9 is feasible, and by increasing SNR to 20, the R2 can increase to a value above 0.99.
+
In Figure 3, we present the scatter plots of simulated K values versus fitted K values using simulated data with different number of diffusion times at various SNR levels. Four cases are provided, including fitting simulated data generated with Δ1 = 19 ms (row 1) or Δ2 = 49 ms (row 2), fitting data with both diffusion times (row 3), and fitting data with three diffusion times (row 4). R2 values for each case at each SNR level are provided. This result verifies that sub-diffusion based kurtosis estimation (blue dots) improves using multiple diffusion times. The improvement in R2 is prominent when moving from fitting single diffusion time data to two diffusion times data, especially when the data is noisy (e.g., SNR = 5 and 10). The improvement gained by moving from two to three distinct diffusion times is marginal (less than 0.01 improvement in R2 value at SNR = 5 and 10, and no improvements for SNR = 20 data). Moreover, our simulation findings highlight the deviation away from the ground truth kurtosis K by using the traditional DKI method (orange dots), especially with kurtosis values larger than 1. Overall, fitting sub-diffusion model (3) to data with two adequately separated diffusion times can lead to robust estimation of mean kurtosis, via (9).
+
+
+Time-dependence in DKI metrics
+
In Figure 4, we show the time-dependence effect of the DKI metrics (DDKI and KDKI) after fitting the standard DKI model to our simulated data. In this fitting, we consider b-values of0, 1000, 1400 and 2500, and vary the diffusion time, as in Jelescu et al. (2022). We depict the parameter estimates, DDKI and KDKI, from simulated data with added noise (SNR = 20) in Figure 4(A) and (B) for the diffusion time between 10-110ms. In Figure 4 (C) and (D), using data with no added noise, we illustrate the long term fitting results and trends in the parameter estimates. In both (A) and (C), as diffusion time increases, DDKI decreases as expected for an effective diffusion coefficient. The mean DDKI (averaged over 1000 simulations) agrees with the ground truth diffusivity D*. When it comes to the kurtosis KDKI, in the noiseless data setting (D), we see KDKI is converging to the ground truth kurtosis value K* at large diffusion time, while in the noisy data setting (B), this trend is not obvious within the experimental diffusion time window. More explanations of the observed time-dependence of diffusivity and kurtosis are provided in the Discussion.
+
+
+
Simulation results on the effect of the number of diffusion times involved in the fitting of the sub-diffusion model (3) parameters (Dβ, β) and computing K* following (9) at various SNR levels. The ground truth values for (Dβ, β, K*) are set to (3 × 10-4, 0.75, 0.8125) to represent white matter (blue) and (5 × 10-4, 0.85, 0.4733) to represent gray matter (orange). Rows 1-3: Distributions of fitted parameter values using different number of diffusion times. 2′ represents an additional simulation using two diffusion times but set to be the same, so it has the same number of data points in the fitting as for using two different diffusion times. Row 4: Coefficient of variation (CV) of the parameter values fitted using different number of diffusion times.
+
+
+
+
+
Surface plots of R2 values achieved with fitting simulated data with two diffusion times, Δ1 and Δ2, to the sub-diffusion model (3) at various SNR levels. R2 values were computed by comparing the estimated mean kurtosis with the ground truth kurtosis. R2 contours at the 0.85, 0.90, 0.95 and 0.99 levels have been provided for visualisation purposes.
+
+
+
+
+Towards fast DKI data acquisitions
+
Next, we sought to identify the minimum number and combination of b-values to use for mean kurtosis estimation based on the sub-diffusion model (3). The simulation results were generated using the Connectome 1.0 DW-MRI data b-value setting, which has 16 b-values, 8 from Δ = 19ms and 8 from Δ = 49ms.
+
In Figure 5, we computed R2 values for all combinations of choices when the number of nonzero b-values is two, three, and four. We then plotted the R2 values sorted in ascending order for each considered SNR level. Notably, as the number of non-zero b-values increase, the achievable combinations increase as well (i.e., 120, 560 and 1820 for the three different non-zero b-value cases). The colours illustrate the proportion of b-values used in the fitting based on Δ1 and Δ2. For example, 1 : 0 means only Δ1 b-values were used, and 2 : 1 means two Δ1 and one Δ2 b-values were used to generate the result. The b-value combinations achieving highest R2 values are displayed in the inset pictures and the b-values are provided in Table 1.
+
Our simulation results in Table 1 suggest that increasing the number of non-zero b-values from two to four improves the quality of the parameter estimation, also achievable by fitting the subdiffusion model (3) to higher SNR data. The gain is larger by using higher SNR data than by using more b-values. For example, going from two to four non-zero b-values with SNR = 5 data approximately doubles the R2, whereas the R2 almost triples when SNR = 5 data is substituted by SNR = 20 data. Additionally, the use of Δ1 or Δ2 alone is not preferred (also see Figure 5), and preference is towards first using Δ1 and then supplementing with b-values generated using Δ2. At all SNR levels when only two non-zero b-values are used, one b-value should be chosen based on the Δ1 set, and the other based on Δ2. Moving to three non-zero b-values requires the addition of another Δ1 b-value, and when four non-zero b-values are used then two from each diffusion time are required. If we consider an R2 = 0.90 to be a reasonable goodness-of-fit for the sub-diffusion model, then at least three or four non-zero b-values are needed with an SNR = 20. If SNR = 10, then three non-zero b-values will not suffice. Interestingly, an R2 of 0.85 can still be achieved when SNR = 20 and two optimally chosen non-zero b-values are used.
+
+
+
Scatter plots of simulated K values vs. fitted K values for simulated data with different number of diffusion times at various SNR levels. The simulated data is created using the sub-diffusion model with random normal noise (10). Blue dots represent kurtosis based on fitting the sub-diffusion model (3). Orange dots represent kurtosis based on fitting the traditional DKI model (6). Black line is a reference line for R2 = 1.00, indicating fitted kurtosis values are 100% matching the simulated ones.
+
+
+
+
+
Time-dependence in DKI metrics using simulated data at different diffusion times (
+
+). The ground truth values for (Dβ, β, K*) used in the simulations are set to (3 × 10-4, 0.75, 0.8125) to represent white matter (blue dotted lines) and (5 × 10-4, 0.85, 0.4733) to represent gray matter (orange dotted lines). (A) and (B) use data with added random Gaussian noise (SNR = 20) to estimate the parameters DDKI and KDKI. (C) and (D) use noiseless data to obtain estimates for large
+
+ values. Shaded regions in (A) and (B) represent the 95% confidence intervals of the estimates.
+
+
+
+
+
R2 values for the b-value sampling optimisation based on DW-MRI data with SNR = 5, 10 and 20. The specifically investigated b-value combinations using two, three and four non-zero b-values have been ordered by the size of the R2 value. The colour bar depicts the proportion of Δ = 19 ms and Δ = 49 ms b-values needed to produce the corresponding R2 value. Note, the different b-value combinations were assigned a unique identifier, and these appear along the abscissa for each of the three non-zero b-value cases. The b-value combinations achieving the highest R2 values are displayed in the inset pictures and the b-values are provided in Table 1.
+
+
+
+
+
A selection of the best b-value sampling regimes to achieve the highest R2 value in the three cases considered. The various categories correspond with two, three and four non-zero b-value sampling schemes, with Δ1 and Δ2 denoting the diffusion time setting used to generate the b-values. Note, entries are b-values in unit of s/mm2, and Δ1 = 19 ms and Δ2 = 49 ms were used to match the Connectome 1.0 DW-MRI data collection protocol. The entries listed at the bottom row are suggested optimal nonzero b-values for clinical practice.
+
+
+
Table 1 summarises findings based on having different number of non-zero b-values with R2 values deduced from the SNR = 5, 10 and 20 simulations. We have chosen to depict five b-value combinations producing the largest R2 values for the two, three and four non-zero b-value sampling cases. We found consistency in b-value combinations across SNR levels. Thus, we can conclude that a range of b-values can be used to achieve a large R2 value, which is a positive finding, since parameter estimation does not stringently rely on b-value sampling. For example, using three non-zero b-values an R2 ≥ 0.90 is achievable based on different b-value sampling. Importantly, two distinct diffusion times are required, and preference is towards including a smaller diffusion time b-value first. Hence, for three non-zero b-values we find two b-values based on Δ1 and one based on Δ2. This finding suggests one of the Δ1 b-values can be chosen in the range 50 s/mm2 to 350 s/mm2, and the other in the range 1500 s/mm2 to 4750 s/mm2. Additionally, the Δ2 b-value can also be chosen in a range, considering between 2300 s/mm2 to 4250 s/mm2 based on the Connectome 1.0 b-value settings. The b-value sampling choices made should nonetheless be in view of the required R2 value. Overall, sampling using two distinct diffusion times appears to provide quite a range of options for the DW-MRI data to be used to fit the sub-diffusion model parameters. The suggested optimal b-value sampling in the last row of Table 1, primarily chosen to minimise b-value size whilst maintaining a large R2 value, may be of use for specific neuroimaging studies, which will be used to inform our discussion on feasibility for clinical practice.
+
+
+Benchmark mean kurtosis in the brain
+
The benchmark mean kurtosis estimation in the brain is established using the entire b-value range with all diffusion encoding directions available in the Connectome 1.0 DW-MRI data. For two subjects in different slices, Figure 6 provides the spatially resolved maps of mean kurtosis computed using the sub-diffusion method (i.e., K*) with one or two diffusion times, and using the standard method (i.e., KDKI) considering the two distinct diffusion times. First, we notice a degradation in the KDKI image with an increase in diffusion time. Second, the use of a single diffusion time with the sub-diffusion model leads to K* values which are larger than either the KDKI values or K* values generated using two diffusion times. Third, the quality of the mean kurtosis map appears to visually be best when two diffusion times are used to estimate K*. Superior grey-white matter tissue contrast (TC) was found for the K* map (TC = 1.73), compared to the KDKI maps (TC = 0.80 for the Δ = 19 ms dataset and TC = 1.01 for the Δ = 49 ms dataset).
+
In Figure 7, an error map (measured by root-mean-square-error, RMSE) from Subject 5 slice 74 was presented for fitting the sub-diffusion model to the DW-MRI data with two diffusion times. Sample parameter fittings in both b-space (3) and q-space (4) were provided for four representative white and grey matter voxels.
+
Quantitative findings are provided in Table 2. The analysis was performed for sub-cortical grey matter (scGM), cortical grey mater (cGM) and white matter (WM) brain regions. For specifics we refer the reader to the appropriate methods section. The table entries highlight the differences in mean kurtosis when computed using the two different approaches. The trend for the traditional single diffusion time approach is that an increase in Δ results in a slight decrease in the mean KDKI, and an increase in the coefficient of variation (CV) for any region. For example, the mean KDKI in the thalamus reduces from 0.65 to 0.58, while the CV increases from 30% to 39%. As much as 30% increase in CV is common for scGM and cGM regions, and around 10% for WM regions. The CV based on the K* value for each region is less than the CV for KDKI with either Δ = 19 ms or Δ = 49 ms.
+
Figure 8 presents the distributions of the fitted parameters (D and K) in specific brain regions, based on the sub-diffusion model (panel A) and the standard DKI model with Δ = 19 ms (panel B). The distributions are colored by the probability density. Yellow indicates high probability density, light blue indicates low probability density. In each subplot, the diffusivity is plotted along the abscissa axis and the kurtosis is along the ordinate axis. Results for the standard DKI model with Δ = 49 ms are qualitatively similar, so are not shown here. From panel (B), we see an unknown nonlinear relationship between the DKI pair, DDKI and KDKI, in all regions considered. By comparison, the sub-diffusion based K* and D* (panel A) are less correlated with each other, indicating D* and K* carry distinct information, which will be very valuable for characterising tissue microstructure.
+
+
+Reduction in DKI data acquisition
+
Results for reductions in diffusion encoding directions to achieve different levels of SNR with the purpose of shortening acquisition times will be benchmarked against the K* maps in Figure 6 and the K* values reported in Table 2.
+
Figure 9 presents the qualitative findings for two subjects generated using all, optimal and sub-optimal b-value samplings with SNR = 6 (3 non-collinear directions, 6 measurements), 10 (8 non-collinear directions, 16 measurements) and 20 (32 non-collinear directions, 64 measurements) DW-MRI data. The quality of the mean kurtosis map improves with increasing SNR, and also by optimising b-value sampling. Optimal sampling at SNR = 10 is qualitatively comparable to the SNR = 20 optimal sampling result, and to the benchmark sub-diffusion results in Figure 6.
+
Quantitative verification of the qualitative observations are provided in Table 3. Significant differences in brain region-specific mean kurtosis values occur at the SNR = 6 level, which are not apparent when SNR = 10 or 20 data with optimal b-value sampling were used. The average errors are relative errors compared to the benchmark kurtosis values reported in Table 2. When using optimal b-values, average errors range from 22% to 43% at SNR = 6, from 13% to 43% at SNR = 10, and from 8% to 20% at SNR = 20, across brain regions. When using sub-optimal b-values, average errors range from 47% to 57% at SNR = 6, from 24% to 102% at SNR = 10, and from 27% to 72% at SNR = 20. Also, the brain region-specific CV for mean kurtosis was not found to change markedly when SNR = 10 or 20 data were used to compute K*. The result of reducing the SNR to 6 leads to an approximate doubling of the CV for each brain region. These findings confirm that with optimal b-value sampling, the data quality can be reduced to around the SNR = 10 level, without a significant impact on the region-specific mean kurtosis estimates derived using the sub-diffusion model.
+
+
+
Spatially resolved maps of mean kurtosis shown for two example slices and two different subjects, Subject 3 rescan slice 71 (Panel A) and Subject 5 slice 74 (Panel B) from the Connectome 1.0 DW-MRI data. Individual maps were generated using the sub-diffusion model framework (K*), as well as using the traditional approach (KDKI). The diffusion times, Δ, used to generate each plot are provided for each case. We consider the mean kurtosis maps using two diffusion times (Δ = 19, 49ms) as the benchmarks.
+
+
+
+
+
Representative error map and sample parameter fits for Subject 5 slice 74. The DW-MRI data with two diffusion times was fitted to the sub-diffusion model in both q-space (A-D) and b-space (E-H), following (3) and (4) respectively. The first and second columns are voxels in white matter (30,20,74) and (45,56,74), respectively. The third and fourth columns are voxels in grey matter (58,35,74) and (34,78,74), respectively.
+
+
+
+
+
Benchmark kurtosis values estimated using the Connectome 1.0 DW-MRI data for different regions of the human brain. Results are provided for the traditional mean kurtosis (KDKI) at two distinct diffusion times, and values (K*) obtained based on fitting the sub-diffusion model across both diffusion times. Results are for grey matter (GM) and white matter (WM) brain regions, in categories of sub-cortical (sc) and cortical (c), and CC stands for corpus callosum. The pooled means and standard deviations across participants have been tabulated, along with the coefficient of variation in parentheses.
+
+
+
+
+
Distributions of the estimated parameter pair (D,K) in different regions of the brain of all subjects, colored by the probability density. Yellow indicates high probability density, light blue indicates low probability density. Panel A: distributions of (D*,K*), generated using the sub-diffusion model (3) with both Δ = 19, 49ms. Panel B: distributions of (DDKI,KDKI), generated using the standard DKI model (6) with Δ = 19ms. Kurtosis is dimensionless and diffusivity is in units of × 10-3 mm2/s.
+
+
+
+
+
Spatially resolved maps of mean kurtosis shown for two example slices and two different subjects, Subject 3 rescan slice 71 (Panel A) and Subject 5 slice 74 (Panel B), based on SNR reduction of the Connectome 1.0 DW-MRI data. Individual maps were generated using the sub-diffusion model framework (K*), considering optimal and sub-optimal four non-zero b-value sampling schemes. Here, two b-values with Δ = 19 ms and two b-values with Δ = 49 ms were selected for each case. The optimal b-values were chosen as the best for each SNR shown in Table 1. The sub-optimal b-values were chosen to have an R2 = 0.3, 0.45, 0.5 to be about half of the maximum R2, for SNR = 6 (b = 800, 1500,200,2300 s/mm2), SNR = 10 (b = 1500, 3450, 6750, 13500 s/mm2) and SNR = 20 (b = 3450, 4750, 2300, 4250 s/mm2), respectively. The benchmark kurtosis map is provided in Figure 6.
+
+
+
+
+Scan-rescan reproducibility of mean kurtosis
+
Figure 10 summarises the intraclass correlation coefficient (ICC) distribution results (μ for mean; σ for standard deviation) for the specific brain regions analysed. The two sets of ICC values were computed based on all DW-MRI (i.e., SNR = 20; subscript A) and the SNR = 10 optimal b-value sampling scheme (subscript O). As the value of μ approaches 1, the inter-subject variation in mean kurtosis is expected to greatly outweigh the intra-subject scan-rescan error. The value of μ should always be above 0.5, otherwise parameter estimation cannot be performed robustly and accurately, and values above 0.75 are generally accepted as good. The μA values for all regions were in the range 0.76 (thalamus) to 0.87 (caudate), and reduced to the range 0.57 (thalamus) to 0.80 (CC) when optimal sampling with SNR = 10 was used to estimate the K* value. Irrespective of which of the two DW-MRI data were used for K* estimation, the value of p. was greater than or equal to 0.70 in 20 out of 24 cases. The μO was less than 0.70 for only the thalamus, putamen and pallidum. The loss in ICC by going to SNR = 10 data with optimal b-value sampling went hand-in-hand with an increase in σ, which is not unexpected, since the uncertainty associated with using less data should be measurable. Overall, μA, μO, and σA, σO, were fairly consistent across the brain regions, suggesting the DW-MRI data with SNR = 10 is sufficient for mean kurtosis estimation based on the sub-diffusion framework.
+
+
+
+Discussion
+
DW-MRI allows the measurement of mean kurtosis, a metric for the deviation away from standard Brownian motion of water in tissue, which has been used to infer variations in tissue microstructure. Research on mean kurtosis has shown benefits in specific applications over other diffusion related measures derived from DW-MRI data (Li et al., 2022b; Liu et al., 2021; Huang et al., 2021a; Guo et al., 2022a; Goryawala et al., 2022; Guo et al., 2022b; Wang et al., 2022; Li et al., 2022a,d; Maralakunte et al., 2022; Hu et al., 2022; Zhou et al., 2022; Li et al., 2022c). Whilst many efforts have been made to optimise mean kurtosis imaging for clinical use, the limitations have been associated with lack of robustness and the time needed to acquire the DW-MRI data for mean kurtosis estimation. The choice of the biophysical model and how diffusion encoding is applied are critical to how well kurtosis in the brain is mapped. Here, we evaluated the mapping of mean kurtosis based on the sub-diffusion model, which allows different diffusion times to be incorporated into the data acquisition. Using simulations and the Connectome 1.0 public DW-MRI dataset, involving a range of diffusion encodings, we showed that mean kurtosis can be mapped robustly and rapidly provided at least two different diffusion times are used and care is taken towards how b-values are chosen given differences in the SNR level of different DW-MRI acquisitions.
+
+Reduction in scan time
+
Previous attempts have been made in optimising the DW-MRI acquisition protocol for mean kurtosis estimation based on the traditional, single diffusion time kurtosis model (Hansen et al., 2013; Hu et al., 2022; Poot et al., 2010; Hansen et al., 2016). Considerations have been made towards reducing the number of b-shells, directions per shell, and sub-sampling of DW-MRI for each direction in each shell. Our findings suggest that robust estimation of kurtosis cannot be achieved using the classical model for mean kurtosis, as highlighted previously (Yang et al., 2022; Ingo et al., 2014, 2015; Barrick et al., 2020). A primary limitation of the traditional method is the use of the cumulant expansion resulting in sampling below a b-value of around 2500 s/mm2 (Jensen etal., 2005; Jensen and Helpern, 2010), and using the sub-diffusion model this limitation is removed (Yang et al., 2022). Our simulation findings and experiments using the Connectome 1.0 data confirm that mean kurtosis can be mapped robustly and rapidly using the sub-diffusion model applied to an optimised DW-MRI protocol. Optimisation of the data acquisition with the use of the sub-diffusion model has not been considered previously.
+
+
+
Kurtosis values (K*) under the optimal and sub-optimal b-value sampling regimes for specific brain regions. K* was estimated based on fitting the sub-diffusion model to the Connectome 1.0 DW-MRI data with two diffusion times and selected four b-shells. Optimal b-value sampling is considered to have R2 = 0.63, 0.91 and 0.96 for the SNR = 6, 10 and 20 columns, according to Table 1. Sub-optimal b-values are chosen to have R2 = 0.3, 0.45 and 0.5, respectively, as reported in Figure 9. Individual entries are for grey matter (GM) and white matter (WM) brain regions, in categories of sub-cortical (sc) and cortical (c), and CC stands for corpus callosum. A reduction in SNR level was achieved by reducing the number of diffusion encoding directions in each b-shell of the DW-MRI data. The pooled means and standard deviations across participants have been tabulated, along with the coefficient of variation in parentheses. The entries identified in italic under the optimal b-value heading were found to be significantly different from the mean K* reported in Table 2. Sub-optimal result population means were mostly significantly different from the mean K*, and they are not italicised. The average errors (last column) are relative errors compared to the benchmark kurtosis values reported in Table 2.
+
+
+
+
+
Interclass correlation coefficient (ICC) results for mean kurtosis are depicted for the 12 brain regions analysed. The mean (μ) and standard deviation (σ) computed based on all the Connectome 1.0 DW-MRI data (A), and the reduced data achieving an SNR = 10 with optimal four non-zero b-value sampling (O), are provided for each brain region. Histograms were generated using all data. Mean kurtosis based on the optimised protocol was computed using the sub-diffusion framework using DW-MRI data with the four non-zero b-values suggested in Table 1 and diffusion encoding directions down sampled to achieve an SNR = 10.
+
+
+
Mean kurtosis values can be generated based on having limited number of diffusion encoding directions (refer to Figure 9 and Table 3). Given that each direction for each b-shell takes a fixed amount of time, then a four b-shell acquisition with six directions per shell will take 25 times longer than a single diffusion encoding data acquisition (assuming a single b-value = 0 data is collected). The total acquisition time for the diffusion MRI protocol for the Connectome 1.0 data was 55 minutes, including 50 b = 0 s/mm2 scans plus seven b-values with 32 and nine with 64 diffusion encoding directions (Tian et al., 2022). This gives a total of 850 scans per dataset. As such, a single 3D image volume took 3.88 s to acquire. Conservatively allowing 4 s per scan, and considering SNR = 20 data (i.e. 64 directions) over four b-values and a single b-value = 0 scan, DW-MRI data for mean kurtosis estimation can be completed in 17 min 8 s (R2 = 0.96). At SNR = 10 (i.e. 32 directions), DW-MRI data with the same number of b-values can be acquired in 4 min 20 s (R2 = 0.91). If an R2 = 0.85 (SNR = 10) is deemed adequate, then one less b-shell is required, saving an additional 64 s. We should point out that even though 2-shell optimised protocols can achieve R2 = 0.85 with SNR = 20, this is not equivalent in time to using 3-shells with SNR = 10 (also R2 = 0.85). This is because 4x additional data are required to double the SNR (equivalent to acquiring an additional 4-shells). However, only one extra b-shell is required to convert 2-shell data to 3-shells with SNR = 10. Our expected DW-MRI data acquisition times are highly feasible clinically, where generally neuroimaging scans take around 15 min involving numerous different MRI contrasts and often a DTI acquisition.
+
An early study on estimating mean kurtosis demonstrated the mapping of a related metric in less than 1 min over the brain (Hansen et al., 2013). Clinical adoption of the protocol lacked, possibly since b-values are a function δ, G and Δ. Hence, different b-values can be obtained using different DW-MRI protocol settings, leading to differences in the mapping of mean kurtosis based on the data (we showed the Δ effect in Figure 6 and Table 2). Our findings suggest this impediment is removed by sampling and fitting data with b-values across two distinct diffusion times. Nonetheless, we should consider what might be an acceptable DW-MRI data acquisition time.
+
A recent study on nasopharyngeal carcinoma investigated reducing the number of b-shell signals based on fixing diffusion encoding directions to the three Cartesian orientations (He et al., 2022). The 3-shell acquisitions took 3 min 2 s to acquire, while the 5-shell data required 5 min 13 s. They investigated as well the impact of using partial Fourier sampling, i.e., reducing the amount of data needed for image reconstruction by reducing k-space line acquisitions for each diffusion encoded image. Their benchmark used 5-shells (200, 400, 800, 1500, 2000 s/mm2), and found partial Fourier sampling with omission of the 1500 s/mm2 b-shell produced acceptable results. With this acquisition the scan could be completed in 3 min 46 s, more than 2× faster than the benchmark 8 min 31 s. Our proposed 3-shells acquisition with an R2 = 0.85 (see SNR = 10 results in Table 1) executable under 4 min is therefore inline with current expectations. Note, at the R2 = 0.85 level the ICC for the different brain regions were in the range 0.60 to 0.69, and these were not formally reported in Figure 10. This level of reproducibility is still acceptable for routine use. We should additionally point out that we used the Subject 1 segmentation labels, after having registered each DW-MRI data to the Subject 1 first scan. This approach results in slight mismatch of the regionspecific segmentations when carried across subjects, inherently resulting in an underestimation of ICC values.
+
Less than 4 min DW-MRI data acquisitions can potentially replace existing data acquisitions used to obtained DTI metrics, since even the estimation of the apparent diffusion coefficient improves by using DW-MRI data relevant to DKI (Veraart et al., 2011b; Wu and Cheung, 2010). Additionally, it is increasingly clear that in certain applications the DKI analysis offers a more comprehensive approach for tissue microstructure analysis (Li et al., 2022b; Liu et al., 2021; Huang et al., 2021a;Guo et al., 2022a; Goryawala et al., 2022; Guo et al., 2022b; Wang et al., 2022; Li et al., 2022a,d; Maralakunte et al., 2022; Hu et al., 2022; Zhou et al., 2022; Li et al., 2022c). As such, multiple b-shell, multiple diffusion encoding direction DW-MRI acquisitions should be used for the calculation of both DTI and DKI metrics.
+
+
+DW-MRI data acquisition considerations
+
To achieve R2 > 0.92 for estimating kurtosis K*, it is necessary to have four b-values, e.g., two relatively small b-values (350 s/mm2 using Δ = 19 ms, and 950 s/mm2 using Δ = 49 ms, both with G = 68 mT/m) and two larger b-values of around 1500 s/mm2 and 4250 s/mm2 (using Δ = 19 ms and Δ = 49 ms respectively, both with G = 142 mT/m) (see bottom row in Table 1). If two or three non-zero b-values are considered sufficient (with R2 = 0.85 or 0.90), then the larger Δ needs to be used to set the largest b-value to be 2300 s/mm2, and the other(s) should be set using the smaller Δ. For the two non-zero b-values case, the b-value from the smaller Δ should be around 800 s/mm2. For the three non-zero b-values case, the b-values from the smaller Δ would then need to be 350 s/mm2 and 1500 s/mm2. Interestingly, b-value = 800 s/mm2 lies around the middle of the 350 s/mm2 to 1500 s/mm2 range. The additional gain to R2 = 0.92 can be achieved by splitting the b-value with the larger Δ into two, again with 2300 s/mm2 near the middle of the two new b-values set. In addition, the separation between Δ1 and Δ2 needs to be as large as plausible, as can be deduced from the simulation result in Figure 2, but attention should be paid to signal-to-noise ratio decreases with increased echo times (He et al., 2022).
+
A recent study on optimising quasi-diffusion imaging (QDI) considered b-values up to 5000 s/mm2 (Spilling et al., 2022). While QDI is a non-Gaussian approach, it is different to the sub-diffusion model, but still uses the Mittag-Leffler function and involves the same number of model parameters. The DW-MRI data used in their study was acquired with a single diffusion time. Nevertheless, particular points are worth noting. Their primary finding was parameter dependence on the maximum b-value used to create the DW-MRI data. They also showed the accuracy, precision and reliability of parameter estimation are improved with increased number of b-shells. They suggested a maximum b-value of 3960 s/mm2 for the 4-shell parameter estimation. Our results do not suggest a dependence of the parameter estimates on maximum b-value (note, if a maximum b-value dependence is present, benchmark versus optimal region specific results in Tables 2 and 3 should show some systematic difference; and we used the real part of the DW-MRI data and not the magnitude as commonly used). These findings potentially confirm that Δ separation is an important component of obtaining a quality parameter fit from which mean kurtosis is deduced (Figure 2).
+
+
+Diffusion gradient pulse amplitudes
+
Commonly available human clinical MRI scanners are capable of 40 mT/m gradient amplitudes. Recently, the increased need to deduce tissue microstructure metrics from DW-MRI measurements has led to hardware developments resulting in 80 mT/m gradient strength MRI scanners. These were initially sought by research centres. The Connectome 1.0 scanners achieve 300 mT/m gradient amplitudes (Tian et al., 2022), which in turn allow large b-values within reasonable echo times. The Connectome 2.0 scanner is planned to achieve 500 mT/m gradient amplitudes (Huang et al., 2021b), providing data for further exploration of the (q, Δ) space by providing a mechanism for increasing our knowledge of the relationships between the micro-, meso- and macro-scales. Whilst there are three Connectome 1.0 scanners available, only one Connectome 2.0 scanner is planned for production. Hence, robust and fast methods utilising existing 80 mT/m gradient systems are needed, and the Connectome scanners provide opportunities for testing and validating methods.
+
The b-value is an intricate combination of δ, G, and Δ. An increase in any of these three parameters increases the b-value (note, δ and G increase b-value quadratically, and Δ linearly). An increase in δ can likely require an increase in Δ, the consequence of which is a reduction in signal-to-noise ratio as the echo time has to be adjusted proportionally. Partial Fourier sampling methods aim to counteract the need to increase the echo time by sub-sampling the DW-MRI data used to generate an image for each diffusion encoding direction (Zong et al., 2021; Heidemann et al., 2010). The ideal scenario, therefore, is to increase G, as for the Connectome 1.0 and 2.0 scanners.
+
Based on our suggested b-value settings in Table 1, a maximum b-value of around 4250 s/mm2 is required for robust mean kurtosis estimation (assume R2 > 0.90 is adequate and achieved via four non-zero b-values and two distinct As, and we should highlight that b-value = 1500 s/mm2 (Δ = 19 ms) and b-value = 4250 s/mm2 (Δ = 49 ms) were achieved using G = 142 mT/m). Considering 80 mT/m gradient sets are the new standard for MRI scanners, an adjustment to δ to compensate for G is needed (recall, q = γδG). Hence, for a constant q, G can be reduced at the consequence of proportionally increasing δ. This then allows MRI systems with lower gradient amplitudes to generate the relevant b-values. For example, changing the δ from 8 ms to 16 ms would result in halving of G (using a maximum G of 142 mT/m from the Connectome 1.0 can achieve the optimal b-values; hence halving would require around 71 mT/m gradient pulse amplitudes). Increasing Δ above 49 ms to create larger b-value data is unlikely to be a viable solution due to longer echo times leading to a loss in SNR. SNR increases afforded by moving from 3 Tto 7 T MRI are most likely counteracted by an almost proportional decrease in T2 times (Pohmann et al., 2016), in addition to 7 T MRI bringing new challenges in terms of increased transmit and static field inhomogeneities leading to signal inconsistencies across an image (Kraff and Quick, 2017).
+
+
+Relationship between sub-diffusion based mean kurtosis <italic>K</italic><sup>*</sup> and histology
+
Following Table 2 (last column), white matter regions showed high kurtosis (0.87±0.22), consistent with a structured heterogeneous environment comprising parallel neuronal fibres, as shown in Maiter et al. (2021). Cortical grey matter showed low kurtosis (0.40±0.16). Subcortical grey matter regions showed intermediate kurtosis (0.60±0.21). In particular, caudate and putamen showed similar kurtosis to grey matter, while thalamus and pallidum showed similar kurtosis properties to white matter. Histological staining results of the subcortical nuclei (Maiter et al., 2021) showed the subcortical grey matter was permeated by small white matter bundles, which could account for the similar kurtosis in thalamus and pallidum to white matter. These results confirmed that our proposed sub-diffusion based mean kurtosis K* is consistent with published histology of normal human brain.
+
+
+Time-dependence of diffusivity and kurtosis
+
The time-dependence of diffusivity and kurtosis has attracted much interest in the field of tissue microstructure imaging. Although our motivation here is not to map time-dependent diffusion, we can nonetheless point out that the assumption ofa sub-diffusion model provides an explanation of the observed time-dependence of diffusivity and kurtosis. From (5), the diffusivity that arises from the sub-diffusion model is of the form
+
+, where
+
+ is the effective diffusion time, DSUB has the standard units for a diffusion coefficient mm2/s, and Dβ is the anomalous diffusion coefficient (with units mm2/sβ) associated with sub-diffusion. In the sub-diffusion framework (1), Dβ and β are assumed to be constant, and hence DSUB exhibits a fractional power-law dependence on diffusion time. Then, following (8), D* is obtained simply by scaling DSUB by a constant 1/Γ(1 + β), and hence also follows a fractional power-law dependence on diffusion time. This time-dependence effect of diffusivity was illustrated in our simulation results, Figure 4(A) and (C).
+
When it comes to kurtosis, the literature on the time-dependence is mixed. Some work showed kurtosis to be increasing with diffusion time in both white and gray matter (Aggarwal et al., 2020), and in gray matter (Ianus et al., 2021), while others showed kurtosis to be decreasing with diffusion time in gray matter (Lee et al., 2020; Olesen et al., 2022; Jelescu et al., 2022). In this study, we provide an explanation of these mixed results. We construct a simulation of diffusion MRI signal data based on the sub-diffusion model (3) augmented with random Gaussian noise. Then we fit the conventional DKI model to the synthetic data. As shown in Figure 4(D), when there is no noise, KDKI increases with diffusion time in white matter, while decreasing with diffusion time in gray matter. When there is added noise, as shown in Figure 4(B), the time-dependency of kurtosis within the timescale of a usual MR experiment is not clear. This goes some way to explaining why the results in the literature on the time-dependence of kurtosis are quite mixed.
+
Furthermore, we summarise the benefits of using the sub-diffusion based mean kurtosis measurement K*. First, as shown in (9), sub-diffusion based mean kurtosis K* is not time-dependent, and hence has the potential to become a tissue-specific imaging biomarker. Second, the fitting of the sub-diffusion model is straightforward, fast and robust, from which the kurtosis K* is simply computed as a function of the sub-diffusion model parameter β, (9). Third, the kurtosis K* is not subject to any restriction on the maximum b-value, as in standard DKI. Hence its value truly reflects the information contained in the full range of b-values.
+
+
+Extension to directional kurtosis
+
The direct link between the sub-diffusion model parameter β and mean kurtosis is well established (Yang et al., 2022; Ingo et al., 2014, 2015). An important aspect to consider is whether mean β used to compute the mean kurtosis is alone sufficient for clinical decision making. While benefits of using kurtosis metrics over other DW-MRI data derived metrics in certain applications are clear, the adequacy of mean kurtosis over axial and radial kurtosis is less apparent. Most studies perform the mapping of mean kurtosis, probably because the DW-MRI data can be acquired in practically feasible times. Nonetheless, we can point to a few recent examples where the measurement of directional kurtosis has clear advantages. A study on mapping tumour response to radiotherapy treatment found axial kurtosis to provide the best sensitivity to treatment response (Goryawala et al., 2022). In a different study a correlation was found between glomerular filtration rate and axial kurtosis is assessing renal function and interstitial fibrosis (Li et al., 2022a). Uniplor depression subjects have been shown to have brain region specific increases in mean and radial kurtosis, while for bipolar depression subjects axial kurtosis decreased in specific brain regions and decreases in radial kurtosis were found in other regions (Maralakunte et al., 2022). This selection of studies highlight future opportunities for extending the methods to additionally map axial and radial kurtosis.
+
Notably, estimates for axial and radial kurtosis require directionality of kurtosis to be resolved, resulting in DW-MRI sampling over a large number of diffusion encoding directions within each b-shell (Jensen and Helpern, 2010; Poot et al., 2010). As such, extension to directional kurtosis requires a larger DW-MRI dataset acquired using an increased number of diffusion encoding directions. The number of b-shells and directions therein necessary for robust and accurate mapping of directional kurtosis based on the sub-diffusion model is an open question.
+
There are three primary ways of determining mean kurtosis. These include the powder averaging over diffusion encoding directions in each shell, and then fitting the model, as in our approach. A different approach is to ensure each b-shell in the DW-MRI data contains the same diffusion encoding directions, and then kurtosis can be estimated for each diffusion encoding direction, after which the average over directions is used to state mean kurtosis. Lastly, the rank-4 kurtosis tensor is estimated from the DW-MRI data, from which mean kurtosis is computed directly. The latter two approaches are potential candidates for extending to axial and radial kurtosis mapping. Note, in DTI a rank-2 diffusion tensor with six unique tensor entries is needed to be estimated, whilst in DKI in addition to the rank-2 diffusion tensor, the kurtosis tensor is rank-4 with 15 unknowns, resulting in 21 unknowns altogether (Hansen et al., 2016). As such, DKI analysis for directional kurtosis requires much greater number of diffusion encoding directions to be sampled than DTI. This automatically means that at least 22 DW-MRI data (including b-value = 0) with different diffusion encoding properties have to be acquired (Jensen and Helpern, 2010). The traditional approach has been to set five distinct b-values with 30 diffusion encoding directions within each b-shell (Poot et al., 2010). Hence, to obtain the entries of the rank-4 kurtosis tensor, much more DW-MRI data is needed in comparison to what is proposed for mean kurtosis estimation in this study. Estimation of the tensor entries from this much data is prone to noise, motion and image artifacts in general (Tabesh et al., 2010), posing challenges on top of long DW-MRI data acquisition times.
+
A kurtosis tensor framework based on the sub-diffusion model where separate diffusion encoding directions are used to fit a direction specific β is potentially an interesting line of investigation for the future, since it can be used to establish a rank-2 β tensor with only six unknowns, requiring at least six distinct diffusion encoding directions. This type of approach can reduce the amount of DW-MRI data to be acquired, and potentially serve as a viable way forward for the combined estimation of mean, axial and radial kurtosis.
+
+
+Kurtosis estimation outside of the brain
+
Although our study has been focusing on mean kurtosis imaging in the human brain, it is clear that DKI has wide application outside of the brain (Li et al., 2022b; Liu et al., 2021; Huang et al., 2021a; Guo et al., 2022a; Li et al., 2022a,d; Zhou et al., 2022). Without having conducted experiments elsewhere, we cannot provide specific guidelines for mean kurtosis imaging in the breast, kidney, liver, and other human body regions. We can, however, point the reader in a specific direction.
+
The classical mono-exponential model can be recovered from the sub-diffusion equation by setting β = 1. For this case, the product between the b-value and fitted diffusivity has been reported to be insightful for b-value sampling (Yablonskiy and Sukstanskii, 2010), in accordance with a theoretical perspective (Istratov and Vyvenko, 1999). It was suggested the product should approximately span the (0, 1) range. Considering our case based on the sub-diffusion equation, we can investigate the size of bDSUB by analysing the four non-zero b-value optimal sampling regime (Δ1: 350 s/mm2 and 1500 s/mm2; Δ2: 950 s/mm2 and 4250 s/mm2 from Table 1). Considering scGM, cGM and WM brain regions alone, the rounded and dimensionless bDSUB values are (0.09, 0.38, 0.20, 0.88), (0.13, 0.55, 0.30, 1.34) and (0.05, 0.23, 0.11, 0.48), respectively, and note that in each case the first two effective sampling values are based on Δ1, and the othertwo are derived using Δ2. Interestingly, the log-linear sampling proposed in (Istratov and Vyvenko, 1999) is closely mimicked by the effective sampling regime (scGM: -2.42, -1.62, -0.97, -0.13; cGM: -2.06, -1.20, -0.60, 0.29; WM: -2.94, -2.23, -1.48, -0.73; by sorting and taking the natural logarithm). This analysis also informs on why it may be difficult to obtain a generally large R2 across the entire brain, since β and DSUB are brain region specific and the most optimal sampling strategy should be β and DSUB specific. Whilst region specific sampling may provide further gains in the R2 value, and improve ICC values for specific brain regions, such data would take a long time to acquire and require extensive post-processing and in-depth analyses.
In biological tissue, the motion of water molecules is hindered by various microstructures, and hence the diffusion can be considerably slower than unhindered, unrestricted, free diffusion of water. The continuous time random walk formalism provides a convenient mathematical framework to model this sub-diffusive behaviour using fractional calculus (Metzler and Klafter, 2000). The resulting probability density function P(x,t) of water molecules at location x (in units of mm) at time t (in units of s) satisfies the time fractional diffusion equation:
+
+
+
+
+where
+
+ is the time fractional derivative of order β (0 < β ≤ 1) in the Caputo sense, Dβ is the generalised anomalous diffusion coefficient with unit of mm2/sβ, and the parameter β characterises the distribution of waiting times between two consecutive steps in the continuous time random walk interpretation. When β = 1, the waiting times have finite mean; when 0 < β < 1, the waiting times have infinite mean, leading to sub-diffusion behaviour. The solution to the time fractional diffusion equation (1) in Fourier space is:
+
+
+
+
+where
+
+ is the single-parameter Mittag-Leffler function, Γ is the standard Gamma function and by definition E1(z) = exp(z). In the context of diffusion DW-MRI, k in (2) represents the q-space parameter q = γGδ, t represents the effective diffusion time
+
+ and p(k, t) represents the signal intensity
+
+, leading to the diffusion signal equation (Magin et al., 2020):
+
+
+
+
+Defining
+
+, the DW-MRI signal then can be expressed in terms of b-values:
+
+
+
+
+where
+
+
+
+
+has the standard unit for a diffusion coefficient, s/mm2.
+
+
+Diffusional kurtosis imaging
+
The traditional DKI approach was proposed by Jensen et al. (Jensen et al., 2005; Jensen and Helpern, 2010) to measure the extent of non-Gaussian diffusion in biological tissues using DW-MRI data:
+
+
+
+
+where S is the signal for a given diffusion weighting b (i.e., b-value), S0 is the signal when b = 0, DDKI and KDKI are the apparent diffusivity and kurtosis. A major limitation of (6) is that it was developed based on the Taylor expansion of the logarithm of the signal at b = 0, as such b-values should be chosen in the neighbourhood of b-value = 0 (Yang et al., 2022; Kiselev, 2010). Hence, to estimate diffusivity and kurtosis, Jensen and Helpern (Jensen and Helpern, 2010) suggested the use of three different b-values (such as o, 1ooo, 2ooo s/mm2) and the maximum b-value should be in the range 2ooo s/mm2 to 3ooo s/mm2 for brain studies. Subsequently, the optimal maximum b-value was found to be dependent on the tissue types and specific pathologies, which makes the experimental design optimal for a whole brain challenging (Chuhutin et al., 2017). The procedure for fitting kurtosis and diffusivity tensors is also not trivial, and a variety of fitting procedures are currently in use. We refer readers to the descriptive and comparative studies for detail on the implementation and comparison of methods (Veraart et al., 2011b,a; Chuhutin et al., 2017).
+
+
+Mean kurtosis from the sub-diffusion model
+
Yang et al. (2022) established that the traditional DKI model corresponds to the first two terms in the expansion of the sub-diffusion model:
+
+
+
+
+where diffusivity, D*, and kurtosis, K*, are computed directly via sub-diffusion parameters DSUB and β:
+
+
+
+
+
+
+
+
+where DSUB is defined in (5). Note the mean kurtosis expression in (9) was also derived by Ingo et al. (Ingo et al., 2015) using a different method. Their derivation follows the definition of kurtosis,
+
+, i.e., by computing the fourth moment
+
+ and the second moment
+
+ based on the sub-diffusion equation (1).
+
+
+
+Connectome 1.0 human brain DW-MRI data
+
The DW-MRI dataset provided by Tian et al. (2022) was used in this study. The publicly available data were collected using the Connectome 1.0 scanner for 26 healthy participants. The first seven subjects had a scan-rescan available. We evaluated qualitatively the seven datasets, and chose the six which had consistent diffusion encoding directions. Subject 2 had 60 instead of 64 diffusion encoding directions, and hence, was omitted from this study. The 2 × 2 × 2 mm3 resolution data were obtained using two diffusion times (Δ = 19, 49 ms) with a pulse duration of δ = 8 ms and G = 31, 68, 105, 142, 179, 216, 253, 290 mT/m, respectively generating b-values = 50,350, 800, 1500, 2400, 3450, 4750, 6000 s/mm2 for Δ = 19 ms, and b-values = 200, 950, 2300, 4250, 6750, 9850, 13500, 17800 s/mm2 for Δ = 49 ms, according to b-value = (γδG)2(Δ - δ/3). 32 diffusion encoding directions were uniformly distributed on a sphere for b < 2400 s/mm2 and 64 uniform directions for b ≤ 2400 s/mm2.
+
The FreeSurfer’s segmentation labels as part of the dataset were used for brain-region specific analyses. Tian et al. (2022) provided the white matter averaged group SNR (23.10 ± 2.46), computed from 50 interspersed b-value = 0 s/mm2 images for each subject. Both magnitude and the real part of the DW-MRI were provided. Based on an in-depth analysis, the use of the real part of the DW-MRI data was recommended, wherein physiological noise, by nature, follows a Gaussian distribution (Gudbjartsson and Patz, 1995).
+
+
+Simulated DW-MRI data at specific b-values
+
DW-MRI data were simulated to establish (i) the correspondence between actual versus fitted mean kurtosis using the traditional DKI and sub-diffusion models based on various choices for Δ, and (ii) to investigate the impact of SNR levels and sub-sampling of b-values on the mean kurtosis estimate. The DW-MRI signal was simulated using the sub-diffusion model (3) with random Gaussian noise added to every normalised DW-MRI signal instance:
+
+
+
+
+where N(0, σ2) is white noise with mean of zero and standard deviation of σ according to the normal distribution.
+
Two aspects influence σ in the case of real-valued DW-MRI data. These include the SNR achieved with a single diffusion encoding direction (i.e., Connectome 1.0 DW-MRI data SNR was derived using only b-value = 0 s/mm2 data), and the number of diffusion encoding directions in each b-shell across which the powder average is computed:
+
+
+
+
+where NDIR is the number of diffusion encoding directions for each b-shell and assuming it is consistent across b-shells. The σ for the Connectome 1.0 data is approximately 1/(23.10×8) = 0.0054 based on 64 diffusion encoding directions. Achieving of SNR = 5, 10 and 20 for the simulation study can therefore be accomplished by changing only the σ and keeping NDIR = 64. As such, σSNR=5 = 0.0250, σSNR=10 = 0.0125 and σSNR=20 = 0.0063.
+
Three simulation experiments were carried out at various SNR levels. The first simulation experiment was to examine the effect of the number of diffusion times on the accuracy of the parameter fitting for idealised grey and white matter cases. In (10) the choices of Dβ = 3 × 10-4 mm2/sβ, β = 0.75, and Dβ = 5 × 10-4 mm2/sβ, β = 0.85, were made for white matter and grey matter, respectively. These two distinct βs led to K* of 0.8125 and 0.4733 using (9). Diffusion times were chosen from the range Δi ∈ [δ, δ + 50], where diffusion pulse length was set to δ = 8 ms to match the Connectome 1.0 data. A minimum required separation between any two Δs was enforced, i.e., 30/(n -1), where 30 corresponds with the 30 ms difference between the Δs for the Connectome 1.0 DW-MRI data, and n is the number of distinct diffusion times simulated. We considered as many as five distinct diffusion times. Simulations were conducted by randomly selecting sets of Δs for 1000 instances, and then generating individual DW-MRI simulated signals using (10), before fitting for Dβ and β, from which K* was computed using (9). For the case of two diffusion times, suggestions on the separation between them was given based on the goodness-of-fit of the model.
+
The second simulation experiment was to investigate the effect of the number of diffusion times on the accuracy of parameter fitting using simulated data with random values of Dβ and β. The domains were restricted to Dβ ∈ [10-4, 10-3] in the unit of mm2/sβ and β ∈ [0.5, 1], corresponding to K* ∈ [0, 1.7124]. For the case of two diffusion times, suggestions on the separation between them were given based on the goodness-of-fit of the model.
+
The third simulation experiment isto study b-value sub-sampling under various SNR levels (SNR = 5, 10 and 20). We set the b-values in the simulated data the same as those used to acquire the Connectome 1.0 dataset. At each SNR level, we selected combinations of two, three and four b-values, irrespective of the difference between them and the diffusion time set to generate the b-value. Essentially, we explored the entire possible sets of b-values for the three regimes, resulting in 120, 560 and 1820 combinations, respectively. A goodness-of-fit measure for model fitting was used to make comparisons between the different b-value combinations.
+
+
+SNR reduction by downsampling diffusion encoding directions
+
We performed SNR reduction of the Connectome 1.0 DW-MRI data by downsampling of diffusion directions in each b-shell. The method of multiple subsets from multiple sets (P-D-MM) subsampling algorithm provided in DMRITool (Cheng et al., 2018) was applied to the b-vectors provided with the Connectome 1.0 DW-MRI data. Note, the b-shells contained 32 directions if b < 2400 s/mm2, and 64 directions if b ≥ 2400 s/mm2. We consider SNR = 20 to be the full dataset. The SNR = 10 data was constructed by downsampling to eight non-collinear diffusion encoding directions in each b-shell, and three were required for the SNR = 6 data. In the downsampled data, each diffusion encoding direction was coupled with the direction of opposite polarity (i.e., SNR = 10 had sixteen measurements for each b value, and SNR = 6 had six).
+
+
+Parameter estimation
+
+Standard DKI model
+
The maximum b-value used to acquire the DW-MRI for standard DKI model fitting is limited to the range 2000 s/mm2 to 3000 s/mm2 due to the quadratic form of (6). We opted to use DW-MRI data generated with b-values = 50, 350, 800, 1500, 2400 s/mm2 using Δ = 19 ms, and b-values = 200, 950, 2300 s/mm2 using Δ = 49 ms. Note, the apparent diffusion coefficient, DDKI in (6) is time dependent, as can be deduced from (5) and (8). Thereby, standard DKI fitting can only be applied to DW-MRI data generated using a single diffusion time. The model in (6) was fitted in a voxelwise manner to the powder averaged (i.e., geometric mean over diffusion encoding directions, often referred to as trace-weighted) DW-MRI data using the lsqcurvefit function in MATLAB (Mathworks, Version 2022a) using the trust-region reflective algorithm. Optimisation function specific parameters were set to TolFun = 10-4 and TolX = 10-6. Parameters were bounded to the ranges of DDKI > 0 and 0 < KDKI ≤ 3.
+
+
+Sub-diffusion model
+
For the single diffusion time case, the sub-diffusion model in (3) was fitted to the powder averaged DW-MRI data in a voxelwise manner using the same MATLAB functions as in the previous section. For each subject, spatially resolved maps of Dβ and β were generated. The fitting strategy for multiple diffusion time data is to solve:
+
+
+
+
+where Sij is the signal at the ith effective diffusion time
+
+, and the jth q-space parameter qj, SUB is the sub-diffusion model (3) at
+
+, and qj for a given set of (Dβ,β), n is the number of diffusion times, and Ji, is the number of q-values corresponding to
+
+, in data acquisition. This objective function allows incorporation of an arbitrary number of diffusion times, each having arbitrary number of q-values. Parameters were bounded to the ranges of 0 < β ≤ 1 and Dβ > 0. Model parameters were found to be insensitive to the choice of initial values. Parameters D* and K* were computed analytically using the estimated Dβ and β according to (8) and (9).
+
+
+
+Goodness-of-fit and region-based statistical analysis
+
The coefficient of determination, referred to as R2, was used to assess how well the mean kurtosis values in the simulation were able to be fitted. It is generally accepted that an R2 value above 0.5 should be achieved and a value of 1.0 is unreasonable for data with realistic noise. Negative values imply the model is a very poor fit. For human data, we computed the region specific mean and standard deviation for each subject, and reported the weighted mean and pooled standard deviation along with the coefficient of variation (CV), defined as the ratio of the standard deviation to the mean. The weights were the number of voxels in the associated regions in each subject. Following Barrick et al. (2020), the tissue contrast is computed as
+
+, where μWM and μGM are the mean parameter values in white and grey matters; and σWM and σGM are the standard deviations of parameter values. Higher TC values indicate greater tissue contrast.
+
Human data were analysed voxelwise, and also based on regions-of-interest. We considered three categories of brain regions, namely sub-cortical grey matter (scGM), cortical grey matter (cGM) and white matter (WM). The scGM region constituted the thalamus (FreeSurfer labels 10 and 49 for left and right hemisphere), caudate (11, 50), putamen (12, 51) and pallidum (13, 52). The cGM region was all regions (1000 to 2999) and separately analysed the fusiform (1007, 2007) and lingual (1013, 2013) brain regions, while the WM had white matter fibre regions from the cerebral (2, 41), cerebellum (7, 46) and corpus callosum (CC; 251 to 255) areas. The average number of voxels in each region were 3986 (scGM), 53326 (cGM), 52121 (WM), 1634 (thalamus), 831 (caudate), 1079 (putamen), 443 (pallidum), 2089 (fusiform), 1422 (lingual), 48770 (cerebral WM), 2904 (cerebellum WM) and 447 (CC). For each brain region a t-test was performed to test for differences in mean kurtosis population means.
+
+
+Scan-rescan analysis using intraclass correlation coefficient (ICC)
+
For each of the six subjects, both the first (scan) and second (rescan) scan images were registered to the first scan images of Subject 1 using inbuilt MATLAB (Mathworks, Version 2022a) functions (imregtform and imwarp). We used 3D affine registration to account for distortions and warps common in DW-MRI data. Cubic spline interpolation was applied to resample both the scan and rescan DW-MRI data for each subject onto Subject 1’s first scan data grid. The FreeSurfer labels for Subject 1’s first scan were used for brain region analysis. The ICC measure was applied to assess scan-rescan reproducibility of mean kurtosis, as described by Duval et al. (Duval et al., 2017) and Fan et al. (Fan et al., 2021). An ICC histogram and the mean and standard deviation descriptive statistics were generated for all brain regions analysed.
+
+
+
+Conclusion
+
The utility of diffusional kurtosis imaging for inferring information on tissue microstructure was described decades ago. Continued investigations in the DW-MRI field have led to studies clearly describing the importance of mean kurtosis mapping to clinical diagnosis, treatment planning and monitoring across a vast range of diseases and disorders. Our research on robust, fast, and accurate mapping of mean kurtosis using the sub-diffusion mathematical framework promises new opportunities for this field by providing a clinically useful, and routinely applicable mechanism for mapping mean kurtosis in the brain. Future studies may derive value from our suggestions and apply methods outside the brain for broader clinical utilisation.
+
+
+
+
+Data and code availability statements
+
The Connectome 1.0 human brain DW-MRI data used in this study is part of the MGH Connectome Diffusion Microstructure Dataset (CDMD)(Tian et al., 2022), which is publicly available on the figshare repository https://doi.org/10.6084/m9.figshare.c.5315474. MATLAB codes generated for simulation study, parameter fitting, and optimising b-value sampling is openly available at https://github.com/m-farquhar/SubdiffusionDKI.
+
+
+Acknowledgement
+
Qianqian Yang and Viktor Vegh acknowledge the financial support from the Australian Research Council (ARC) Discovery Project scheme (DP190101889) for funding a project on mathematical model development and MRI-based investigations into tissue microstructure in the human brain. Qianqian Yang also acknowledges the support from the ARC Discovery Early Career Research Award (DE150101842) for funding a project on new mathematical models for capturing heterogeneity in human brain tissue. Authors also thank the members of the Anomalous Relaxation and Diffusion Study (ARDS) group for many interesting discussions involving diffusion MRI.
+
+
+AggarwalM, SmithMD, CalabresiPA. Diffusion-time dependence of diffusional kurtosis in the mouse brain. Magnetic Resonance in Medicine. 2020Feb; 84(). doi: 10.1002/mrm.28189.
+BarrickTR, SpillingCA, IngoC, MadiganJ, IsaacsJD, RichP, JonesTL, MaginRL, HallMG, HoweFA. Quasidiffusion magnetic resonance imaging (QDI): A fast, high b-value diffusion imaging technique. NeuroImage. 2020may; 211:116606. doi: 10.1016/j.neuroimage.2020.116606.
+ChengJ, ShenD, YapPT, BasserPJ. Single- and multiple-shell uniform sampling schemes for diffusion MRI using spherical codes. IEEE Transactions on Medical Imaging. 2018jan; 37():185-199. doi: 10.1109/tmi.2017.2756072.
+ChuhutinA, HansenB, JespersenSN. Precision and accuracy of diffusion kurtosis estimation and the influence of b-value selection. NMR in Biomedicine. 2017aug; 30():e3777. doi: 10.1002/nbm.3777.
+DuvalT, SmithV, StikovN, KlawiterEC, Cohen-AdadJ. Scan-rescan of axcaliber, macromolecular tissue volume, and g-ratio in the spinal cord. Magnetic Resonance in Medicine. 2017oct; 79():2759-2765. doi: 10.1002/mrm.26945.
+FanQ, PolackalMN, TianQ, NgamsombatC, NummenmaaA, WitzelT, KlawiterEC, HuangSY. Scan-rescan repeatability of axonal imaging metrics using high-gradient diffusion MRI and statistical implications for study design. NeuroImage. 2021oct; 240:118323. doi: 10.1016/j.neuroimage.2021.118323.
+GoryawalaM, MellonEA, ShimH, MaudsleyAA. Mapping early tumor response to radiotherapy using diffusion kurtosis imaging. The Neuroradiology Journal. 2022aug; p. 197140092211222. doi: 10.1177/19714009221122204.
+GudbjartssonH, PatzS. The Rician distribution of noisy MRI data. Magnetic Resonance in Medicine. 1995dec; 34():910-914. doi: 10.1002/mrm.1910340618.
+GuoJ, SunW, DongC, WuZ, LiX, ZhouR, XuW. Intravoxel incoherent motion imaging combined with diffusion kurtosis imaging to assess the response to radiotherapy in a rabbit VX2 malignant bone tumor model. Cancer Imaging. 2022sep; 22(). doi: 10.1186/s40644-022-00488-w.
+GuoM, ShenB, LiJ, HuangX, HuJ, WeiX, WangS, YuanR, HeC, LiY. Diffusion abnormality in temporal lobe epilepsy patients with sleep disorders: A diffusion kurtosis imaging study. Frontiers in Psychiatry. 2022may; 13. doi: 10.3389/fpsyt.2022.885477.
+HansenB, LundTE, SangillR, JespersenSN. Experimentally and computationally fast method for estimation of a mean kurtosis. Magnetic Resonance in Medicine. 2013apr; 69():1754-1760. doi: 10.1002/mrm.24743.
+HansenB, ShemeshN, JespersenSN. Fast imaging of mean, axial and radial diffusion kurtosis. Neuroimage. 2016nov; 142:381-393. doi: 10.1016/j.neuroimage.2016.08.022.
+HeY, ChenH, ZhangH, GrimmR, ZhaoC, GuoX, LiuY, YuanZ. Optimization of scan parameters to reduce acquisition time for RESOLVE-based diffusion kurtosis imaging (DKI) in nasopharyngeal carcinoma (NPC). The British Journal of Radiology. 2022aug; 95(). doi: 10.1259/bjr.20210641.
+HeidemannRM, PorterDA, AnwanderA, FeiweierT, HeberleinK, KnöscheTR, TurnerR. Diffusion imaging in humans at 7T using readout-segmented EPI and GRAPPA. Magnetic Resonance in Medicine. 2010jun; 64():9-14. doi: 10.1002/mrm.22480.
+HenriquesRN, JespersenSN, JonesDK, VeraartJ. Toward more robust and reproducible diffusion kurtosis imaging. Magnetic Resonance in Medicine. 2021apr; 86():1600-1613. doi: 10.1002/mrm.28730.
+HuR, KimH, KimJ, AllenJW, SunPZ. Fast diffusion kurtosis imaging in acute ischemic stroke shows mean kurtosis-diffusivity mismatch. Journal of Neuroimaging. 2022apr; 32():941-946. doi: 10.1111/jon.13000.
+HuangN, ChenY, SheD, XingZ, ChenT, CaoD. Diffusion kurtosis imaging and dynamic contrast-enhanced MRI for the differentiation of parotid gland tumors. European Radiology. 2021oct; 32():2748-2759. doi: 10.1007/s00330-021-08312-y.
+HuangSY, WitzelT, KeilB, ScholzA, DavidsM, DietzP, RummertE, RambR, KirschJE, YendikiA, FanQ, TianQ, Ramos-LlordenG, LeeHH, NummenmaaA, BilgicB, SetsompopK, WangF, AvramAV, KomloshM, et al.Connectome 2.0: Developing the next-generation ultra-high gradient strength human MRI scanner for bridging studies of the micro-, meso- and macro-connectome. NeuroImage. 2021nov; 243:118530. doi: 10.1016/j.neuroimage.2021.118530.
+IanusA, AlexanderDC, ZhangH, PalomboM. Mapping complex cell morphology in the grey matter with double diffusion encoding MR: A simulation study. NeuroImage. 2021; 241. doi: 10.1016/j.neuroimage.2021.118424.
+IngoC, MaginR, ParrishT. New insights into the fractional order diffusion equation using entropy and kurtosis. Entropy. 2014nov; 16():5838-5852. doi: 10.3390/e16115838.
+IngoC, SuiY, ChenY, ParrishTB, WebbAG, RonenI. Parsimonious continuous time random walk models and kurtosis for diffusion in magnetic resonance of biological tissue. Frontiers in Physics. 2015mar; 3:11. doi: 10.3389/fphy.2015.00011.
+IstratovAA, VyvenkoOF. Exponential analysis in physical phenomena. Review of Scientific Instruments. 1999feb; 70():1233-1257. doi: 10.1063/1.1149581.
+JelescuIO, de SkowronskiA, GeffroyF, PalomboM, NovikovDS. Neurite Exchange Imaging (NEXI): A minimal model of diffusion in gray matter with inter-compartment water exchange. NeuroImage. 2022aug; 256:119277. doi: 10.1016/j.neuroimage.2022.119277.
+JensenJH, HelpernJA. MRI quantification of non-Gaussian water diffusion by kurtosis analysis. NMR in Biomedicine. 2010may; 23():698-710. doi: 10.1002/nbm.1518.
+JensenJH, HelpernJA, RamaniA, LuH, KaczynskiK. Diffusional kurtosis imaging: The quantification of nonGaussian water diffusion by means of magnetic resonance imaging. Magnetic Resonance in Medicine. 2005; 53():1432-1440. doi: 10.1002/mrm.20508.
+KiselevVG. The cumulant expansion: An overarching mathematical framework for under standing diffusion NMR. In: Diffusion MRIOxford University Press; 2010.p. 152-168. doi: 10.1093/med/9780195369779.003.0010.
+KraffO, QuickHH. 7T: Physics, safety, and potential clinical applications. Journal of Magnetic Resonance Imaging. 2017apr; 46():1573-1589. doi: 10.1002/jmri.25723.
+KuderTA, StieltjesB, BachertP, SemmlerW, LaunFB. Advanced fit of the diffusion kurtosis tensor by directional weighting and regularization. Magnetic Resonance in Medicine. 2011aug; 67():1401-1411. doi: 10.1002/mrm.23133.
+Le BihanD, Johansen-BergH. Diffusion MRI at 25: Exploring brain tissue structure and function. NeuroImage. 2012jun; 61():324-341. doi: 10.1016/j.neuroimage.2011.11.006.
+Le BihanD, ManginJF, PouponC, ClarkCA, PappataS, MolkoN, ChabriatH. Diffusion tensor imaging: Concepts and applications. Journal of Magnetic Resonance Imaging. 2001; 13():534-546. doi: 10.1002/jmri.1076.
+LebelC, TreitS, BeaulieuC. A review of diffusion MRI of typical white matter development from early childhood to young adulthood. NMR in Biomedicine. 2019; 32():e3778. doi: 10.1002/nbm.3778.
+LeeHH, PapaioannouA, NovikovDS, FieremansE. In vivo observation and biophysical interpretation of time-dependent diffusion in human cortical gray matter. NeuroImage. 2020; 222. doi: 10.1016/j.neuroimage.2020.117054.
+LiA, YuanG, HuY, ShenY, HuX, HuD, LiZ. Renal functional and interstitial fibrotic assessment with nonGaussian diffusion kurtosis imaging. Insights into Imaging. 2022apr; 13(). doi: 10.1186/s13244-022-01215-6.
+LiHW, YanGW, YangJ, ZhuoLH, BhetuwalA, LongYJ, FengX, YaoHC, ZouXX, FengRH, YangHF, DuY. Quantitative analysis for detection and grading of hepatocellular carcinoma: Comparison of diffusion kurtosis imaging, intravoxel incoherent motion and conventional diffusion-weighted imaging. Oncology Letters. 2022sep; 24(). doi: 10.3892/ol.2022.13523.
+LiQ, CaoJ, LiuX, LuoX, SuG, WangD, LinB. The diagnostic value of diffusion kurtosis imaging in Parkinson’s disease: a systematic review and meta-analysis. Annals of Translational Medicine. 2022apr; 10():474-474. doi: 10.21037/atm-22-1461.
+LiQ, CaoB, LiuK, SunH, DingY, YanC, WuPY, DaiC, RaoS, ZengM, JiangS, ZhouJ. Detecting the muscle invasiveness of bladder cancer: An application of diffusion kurtosis imaging and tumor contact length. European Journal of Radiology. 2022jun; 151:110329. doi: 10.1016/j.ejrad.2022.110329.
+LiuY, ZhangGMY, PengX, LiX, SunH, ChenL. Diffusion kurtosis imaging as an imaging biomarker for predicting prognosis in chronic kidney disease patients. Nephrology Dialysis Transplantation. 2021jul; 37():1451-1460. doi: 10.1093/ndt/gfab229.
+MaginRL, HallMG, KaramanMM, VeghV. Fractional calculus models of magnetic resonance phenomena: Relaxation and diffusion. Critical Reviews in Biomedical Engineering. 2020; 48():285-326. doi: 10.1615/critrevbiomedeng.2020033925.
+MaiterA, RiemerF, AllinsonK, ZaccagnaF, Crispin-OrtuzarM, GehrungM, McLeanMA, PriestAN, GristJ, MatysT, GravesMJ, GallagherFA. Investigating the relationship between diffusion kurtosis tensor imaging (DKTI) and histology within the normal human brain. Scientific Reports. 2021; 11. doi: 10.1038/s41598-021-87857-w.
+MaralakunteM, GuptaV, GroverS, AhujaCK, SahooS, KishoreK, VyasS, GargG, SinghP, GovindV. Crosssectional analysis of whole-brain microstructural changes in adult patients with bipolar and unipolar depression by diffusion kurtosis imaging. The Neuroradiology Journal. 2022jul; p. 197140092211144. doi: 10.1177/19714009221114446.
+MetzlerR, KlafterJ. The random walk’s guide to anomalous diffusion: a fractional dynamics approach. Physics Reports. 2000dec; 339():1-77. doi: 10.1016/s0370-1573(00)00070-3.
+OlesenJL, ØstergaardL, ShemeshN, JespersenSN. Diffusion time dependence, power-law scaling, and exchange in gray matter. NeuroImage. 2022may; 251:118976. doi: 10.1016/j.neuroimage.2022.118976.
+PohmannR, SpeckO, SchefflerK. Signal-to-noise ratio and MR tissue parameters in human brain imaging at 3, 7, and 9.4Tesla using current receive coil arrays. Magnetic Resonance in Medicine. 2016mar; 75():801-809. doi: 10.1002/mrm.25677.
+PootDHJ, den DekkerAJ, AchtenE, VerhoyeM, SijbersJ. Optimal experimental design for diffusion kurtosis imaging. IEEE Transactions on Medical Imaging. 2010mar; 29():819-829. doi: 10.1109/tmi.2009.2037915.
+ShaftoMA, TylerLK, DixonM, TaylorJR, RoweJB, CusackR, CalderAJ, Marslen-WilsonWD, DuncanJ, DalgleishT, HensonRN, BrayneC, MatthewsFE. The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study protocol: a cross-sectional, lifespan, multidisciplinary examination of healthy cognitive ageing. BMC Neurology. 2014oct; 14(). doi: 10.1186/s12883-014-0204-1.
+SpillingCA, HoweFA, BarrickTR. Optimization of quasi-diffusion magnetic resonance imaging for quantitative accuracy and time-efficient acquisition. Magnetic Resonance in Medicine. 2022aug; 88():2532-2547. doi: 10.1002/mrm.29420.
+TabeshA, JensenJH, ArdekaniBA, HelpernJA. Estimation of tensors and tensor-derived measures in diffusional kurtosis imaging. Magnetic Resonance in Medicine. 2010oct; 65():823-836. doi: 10.1002/mrm.22655.
+TianQ, FanQ, WitzelT, PolackalMN, OhringerNA, NgamsombatC, RussoAW, MachadoN, BrewerK, WangF, SetsompopK, PolimeniJR, KeilB, WaldLL, RosenBR, KlawiterEC, NummenmaaA, HuangSY. Comprehensive diffusion MRI dataset for in vivo human brain microstructure mapping using 300 mT/m gradients. Scientific Data. 2022jan; 9(). doi: 10.1038/s41597-021-01092-6.
+TournierJD. Diffusion MRI in the brain-Theory and concepts. Progress in Nuclear Magnetic Resonance Spectroscopy. 2019jun; 112-113:1-16. doi: 10.1016/j.pnmrs.2019.03.001.
+Van EssenDC, SmithSM, BarchDM, BehrensTEJ, YacoubE, UgurbilK. The WU-Minn Human Connectome Project: An overview. NeuroImage. 2013oct; 80:62-79. doi: 10.1016/j.neuroimage.2013.05.041.
+VeraartJ, HeckeWV, SijbersJ. Constrained maximum likelihood estimation of the diffusion kurtosis tensor using a Rician noise model. Magnetic Resonance in Medicine. 2011mar; 66():678-686. doi: 10.1002/mrm.22835.
+VeraartJ, PootDHJ, HeckeWV, BlockxI, der LindenAV, VerhoyeM, SijbersJ. More accurate estimation of diffusion tensor parameters using diffusion kurtosis imaging. Magnetic Resonance in Medicine. 2011; 65():138-145. doi: 10.1002/mrm.22603.
+WangML, WeiXE, YuMM, LiWB. Cognitive impairment in mild traumatic brain injury: a diffusion kurtosis imaging and volumetric study. Acta Radiologica. 2022feb; 63():504-512. doi: 10.1177/0284185121998317.
+WuEX, CheungMM. MR diffusion kurtosis imaging for neural tissue characterization. NMR in Biomedicine. 2010jul; 23():836-848. doi: 10.1002/nbm.1506.
+YablonskiyDA, SukstanskiiAL. Theoretical models of the diffusion weighted MR signal. NMR in Biomedicine. 2010jun; 23():661-681. doi: 10.1002/nbm.1520.
+YangQ, ReutensDC, VeghV. Generalisation of continuous time random walk to anomalous diffusion MRI models with an age-related evaluation of human corpus callosum. NeuroImage. 2022jan; p. 118903. doi: 10.1016/j.neuroimage.2022.118903.
+ZelinskiAC, AngeloneLM, GoyalVK, BonmassarG, AdalsteinssonE, WaldLL. Specific absorption rate studies of the parallel transmission of inner-volume excitations at 7T. Journal of Magnetic Resonance Imaging. 2008oct; 28():1005-1018. doi: 10.1002/jmri.21548.
+ZhouZ, ChenY, ZhaoF, SunZ, ZhuL, YuH, WangW. Predictive value of intravoxel incoherent motion combined with diffusion kurtosis imaging for breast cancer axillary lymph node metastasis: a retrospective study. Acta Radiologica. 2022jun; p. 028418512211076. doi: 10.1177/02841851221107626.
+ZongF, DuJ, DengX, ChaiX, ZhuoY, VeghAV, XueR. Fast diffusion kurtosis mapping of human brain at 7 Tesla with hybrid principal component analyses. IEEE Access. 2021; 9:107965-107975. doi: 10.1109/ac-cess.2021.3100546.
+
+
+